The DOI matches its Crossref record; 71 of the 88 references it cites were found in a registry or an index; 8 of the 27 sources its claims cite were checked for support.
Vision Transformers for Dense Prediction
Open accessAbstract
We introduce dense prediction transformers, an architecture that leverages vision transformers in place of convolutional networks as a backbone for dense prediction tasks. We assemble tokens from various stages of the vision transformer into image-like representations at various resolutions and progressively combine them into full-resolution predictions using a convolutional decoder. The transformer backbone processes representations at a constant and relatively high resolution and has a global receptive field at every stage. These properties allow the dense prediction transformer to provide finer-grained and more globally coherent predictions when compared to fully-convolutional networks. Our experiments show that this architecture yields substantial improvements on dense prediction tasks, especially when a large amount of training data is available. For monocular depth estimation, we observe an improvement of up to 28% in relative performance when compared to a state-of-the-art fully-convolutional network. When applied to semantic segmentation, dense prediction transformers set a new state of the art on ADE20K with 49.02% mIoU. We further show that the architecture can be fine-tuned on smaller datasets such as NYUv2, KITTI, and Pascal Context where it also sets the new state of the art. Our models are available at https://github.com/intel-isl/DPT.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
32 Own result · 23 Own interpretation · 11 No basis given · 10 No full text of the sources · 3 Supported · 2 Partly supported · 1 Supported so far · 1 Own method · accuracy not yet measured
Every source the paper cites for this claim was checked and supports it.
In the paper
“Specifically, we use the recently proposed vision transformer (ViT) [11] as a backbone architecture.”
§ Vision Transformers for Dense Prediction > 1. Introduction
What the sources say
- Supportedcites [11] 11. Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. source
Section 3.1 (p. 3) introduces and describes the Vision Transformer (ViT) architecture. The citation identifies the ViT used as the citing authors’ backbone; their choice to use it is not a claim attributed to this source. (LLM, conf=0.93)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Virtually all existing architectures for dense prediction are based on convolutional networks [6,31,34,42,49,50,53].”
§ Vision Transformers for Dense Prediction > 1. Introduction
What the sources say
- Partly supportedcites [6] 6. Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos, Kevin Murphy, and Alan L. Yuille. DeepLab: Semantic im- age segmentation with deep convolutional nets, atrous con- volution, and fully connected crfs. TPAMI, 40(4):834-848, 2018. sourceExperimental — accuracy not yet measured
The Introduction (p. 1) says convolutional models such as PixelCNN had received much more attention for pixel-by-pixel image generation. It does not establish that virtually all architectures across the broader category of dense prediction are convolutional; it also discusses recurrent models and proposes a self-attention alternative (Introduction, p. 2). (LLM, conf=0.78)
- Not checked — no full text of the sourcecites [31] 31. Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U- Net: Convolutional networks for biomedical image segmen- tation. In MICCAI, 2015. source
- Supportedcites [34] 34. Evan Shelhamer, Jonathan Long, and Trevor Darrell. Fully convolutional networks for semantic segmentation. CVPR, 2015. source
The paper is an example of a convolutional-network architecture for dense prediction: its Abstract (p. 1) describes predicting a depth map, and §3.1.2 (p. 3) identifies its convolutional layers and target-depth output. As one of several citations, it supports the sentence by exemplifying such an architecture, rather than establishing the prevalence claim on its own. (LLM, conf=0.88)
- Could not verifycites [42] 42. Jingdong Wang, Ke Sun, Tianheng Cheng, Borui Jiang, Chaorui Deng, Yang Zhao, Dong Liu, Yadong Mu, Mingkui Tan, Xinggang Wang, Wenyu Liu, and Bin Xiao. Deep high-resolution representation learning for visual recogni- tion. TPAMI, 2020. sourceExperimental — accuracy not yet measured
Section 2.1 (p. 8) describes transformer language-model architectures, not architectures for dense prediction. Searches for dense prediction, convolutional networks, and pixel-wise segmentation found no passage addressing whether dense-prediction architectures are predominantly convolutional. (LLM, conf=0.78)
- Not matchedcites [49] 49. Fisher Yu and Vladlen Koltun. Multi-scale context aggrega- tion by dilated convolutions. In ICLR, 2016.
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not checked — no full text of the sourcecites [50] 50. Yuhui Yuan, Xilin Chen, and Jingdong Wang. Object- contextual representations for semantic segmentation. In ECCV, 2020. source
- Not checked — no full text of the sourcecites [53] 53. Hengshuang Zhao, Jianping Shi, Xiaojuan Qi, Xiaogang Wang, and Jiaya Jia. Pyramid scene parsing network. In CVPR, 2017. source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“We notice that the biggest gains in performance for zero-shot transfer were achieved for datasets that feature dense, high-resolution evaluations [15,55,59].”
§ Vision Transformers for Dense Prediction > Additional information
What the sources say
- Not checked — no full text of the sourcecites [15] 15. Andreas Geiger, Philip Lenz, and Raquel Urtasun. Are we ready for autonomous driving? The KITTI vision benchmark suite. In CVPR, 2012. source
- Supportedcites [55] 55. D. J. Butler, J. Wulff, G. B. Stanley, and M. J. Black. A naturalistic open source movie for optical flow evaluation. In ECCV, 2012. source
The citation identifies MPI-Sintel as one of the datasets in the citing authors’ comparison; the performance gains are the citing authors’ observation, not a result attributed to this paper. Design Decisions and Comparison to Middlebury (p. 5) describes Sintel’s higher-resolution frames, and The Sintel Data Set (p. 9) describes its flow fields and pixel-level error evaluation. (LLM, conf=0.86)
- Not checked — no full text of the sourcecites [59] 59. Thomas Schöps, Johannes L. Schönberger, Silvano Galliani, Torsten Sattler, Konrad Schindler, Marc Pollefeys, and An- dreas Geiger. A multi-view stereo benchmark with high- resolution images and multi-camera videos. In CVPR, 2017. source
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Note that ResNext101-WSL was pretrained on a billion-scale corpus of weakly supervised data [25] in addition to ImageNet pretraining.”
§ Vision Transformers for Dense Prediction > Skip connections.
What the sources say
- Partly supportedcites [25] 25. Dhruv Mahajan, Ross Girshick, Vignesh Ramanathan, Kaiming He, Manohar Paluri, Yixuan Li, Ashwin Bharambe, and Laurens van der Maaten. Exploring the limits of weakly supervised pretraining. In ECCV, 2018. sourceExperimental — accuracy not yet measured
Section 2.3 (p. 4) identifies the models as ResNeXt-101, and Section A.5 (p. 23) says they were pretrained on billions of Instagram images using hashtags as labels. But it describes subsequent training on ImageNet as fine-tuning, not additional ImageNet pretraining; Section 3.2 (p. 12) also distinguishes Instagram-pretrained models from ImageNet-pretrained baselines. (LLM, conf=0.88)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“While ignoring the token yields good performance, projection provides slightly better performance on average.”
§ Vision Transformers for Dense Prediction > Skip connections.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Adding the token, on the other hand, yields worse performance than simply ignoring it.”
§ Vision Transformers for Dense Prediction > Skip connections.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Both DPT variants significantly outperform the state of the art.”
§ Vision Transformers for Dense Prediction > 4.1. Monocular Depth Estimation
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Note that both architectures have similar latency to MiDaS (Table 9).”
§ Vision Transformers for Dense Prediction > 4.1. Monocular Depth Estimation
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“ViT-Large outperforms all other backbones but is also almost three times larger than ViT-Base and ViT-Hybrid.”
§ Vision Transformers for Dense Prediction > Skip connections.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We again see that DPT can provide strong performance even on smaller datasets.”
§ Vision Transformers for Dense Prediction > ADE20K.
plainSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Probably not a substantial claim. 11
- It may state two separate claims as one. 13
- The quote only partly states this claim. 1
Every source the paper cites for this claim was checked and supports it.
In the paper
“In particular, it has recently been demonstrated that a direct application of token-based transformer architectures that have been successful in NLP can yield competitive performance on image classification [11].”
§ Vision Transformers for Dense Prediction > 2. Related Work
What the sources say
- Supportedcites [11] 11. Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. source
The Introduction (p. 1) says image patches are treated like NLP tokens and fed to a standard Transformer. The Conclusion (p. 9) reports that this approach matches or exceeds the state of the art on many image-classification datasets when coupled with pre-training on large datasets. (LLM, conf=0.93)
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“A key insight of this work was that, like transformer models in NLP, vision transformers need to be paired with a sufficient amount of training data to realize their potential.”
§ Vision Transformers for Dense Prediction > 2. Related Work
What the sources say
- Supportedcites [11] 11. Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. source
The Introduction (p. 1) says ViT performs modestly on mid-sized datasets and does not generalize well with insufficient data, but attains excellent results when pre-trained at sufficient scale. Section 4.3 (p. 6) reports that larger ViT models show their full benefit only with the largest pre-training dataset tested. The Conclusion (p. 9) also describes the approach as working well when coupled with pre-training on large datasets. (LLM, conf=0.94)
plainconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Supported
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“Similarly, the ViT-Hybrid architecture extracts features at 1 16 the input resolution, which is twice as high as the lowest-resolution features that are commonly used with convolutional backbones.”
§ Vision Transformers for Dense Prediction > 3. Architecture
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds. It may state two separate claims as one.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“It contains about 1.4 million images and is, to the best of our knowledge, the largest training set for monocular depth estimation that has ever been compiled.”
§ Vision Transformers for Dense Prediction > 4.1. Monocular Depth Estimation
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“When applied to semantic segmentation, dense vision transformers set a new state of the art on ADE20K with 49.02% mIoU.”
§ Vision Transformers for Dense Prediction > Abstract
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The quote only partly states this claim.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We observe that the DPT tends to produce cleaner and finer-grained delineations of object boundaries and that the predictions are also in some cases less cluttered.”
§ Vision Transformers for Dense Prediction > ADE20K.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“DPT can better reconstruct fine details while also improving global coherence in areas that are challenging for the convolutional architecture (for example, large homogeneous regions or relative depth arrangement across the image).”
§ Vision Transformers for Dense Prediction > 4.1. Monocular Depth Estimation
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The average relative improvement over the best published architecture, MiDaS, is more than 23% for DPT-Hybrid and 28% for DPT-Large.”
§ Vision Transformers for Dense Prediction > 4.1. Monocular Depth Estimation
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“DPT-Hybrid achieves this with a comparable network capacity (Table 9), while DPT-Large is about 3 times larger than MiDaS.”
§ Vision Transformers for Dense Prediction > 4.1. Monocular Depth Estimation
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We observe the tendency that attention is spatially more localized close to the reference token in shallow layers (leftmost columns), whereas deeper layers (rightmost columns) frequently attend across the whole image.”
§ Vision Transformers for Dense Prediction > Additional information
hedgedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“ViT-Base has comparable performance to ResNext101-WSL, while ViT-Hybrid and ViT-Large improve performance even though they have been pretrained on significantly less data.”
§ Vision Transformers for Dense Prediction > Skip connections.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
In the paper
“While we observe a general trend of an improvement in per-class IoU in comparison to the baseline [51], we do not observe a strong pattern across classes.”
§ Vision Transformers for Dense Prediction > Additional information
What the sources say
- Not checked — no full text of the sourcecites [51] 51. Hang Zhang, Chongruo Wu, Zhongyue Zhang, Yi Zhu, Zhi Zhang, Haibin Lin, Yue Sun, Tong He, Jonas Muller, R. Manmatha, Mu Li, and Alexander Smola. ResNeSt: Split- attention networks. arXiv preprint arXiv:2004.08955, 2020.
hedgedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“While we observe a general trend of an improvement in per-class IoU in comparison to the baseline [51], we do not observe a strong pattern across classes.”
§ Vision Transformers for Dense Prediction > Additional information
What the sources say
- Not checked — no full text of the sourcecites [51] 51. Hang Zhang, Chongruo Wu, Zhongyue Zhang, Yi Zhu, Zhi Zhang, Haibin Lin, Yue Sun, Tong He, Jonas Muller, R. Manmatha, Mu Li, and Alexander Smola. ResNeSt: Split- attention networks. arXiv preprint arXiv:2004.08955, 2020.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We observe more details and also better global depth arrangement in DPT predictions when compared to the fullyconvolutional baseline.”
§ Vision Transformers for Dense Prediction > Additional information
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The transformer backbone processes representations at a constant and relatively high resolution and has a global receptive field at every stage.”
§ Vision Transformers for Dense Prediction > Abstract
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: unclear — It is unclear why the paper says it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“As long as the image size is divisible by p, the embedding procedure can be applied and will produce a varying number of image tokens N p .”
§ Vision Transformers for Dense Prediction > 3. Architecture
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Unlike fully-convolutional networks, the vision transformer backbone foregoes explicit downsampling operations after an initial image embedding has been computed and maintains a representation with constant dimensionality throughout all processing stages.”
§ Vision Transformers for Dense Prediction > 1. Introduction
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Since transformers are known to realize their full potential only when an abundance of training data is available, monocular depth estimation is an ideal task to test the capabilities of DPT.”
§ Vision Transformers for Dense Prediction > 4.1. Monocular Depth Estimation
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“This is in stark contrast to convolutional networks, which progressively increase their receptive field as features pass through consecutive convolution and downsampling layers.”
§ Vision Transformers for Dense Prediction > 3. Architecture
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper may give no reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“DPT-Large performs slightly worse, likely because of the significantly smaller dataset compared to our previous experiments.”
§ Vision Transformers for Dense Prediction > ADE20K.
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“DeIT trains the ViT architecture with a more data-efficient pretraining procedure.”
§ Vision Transformers for Dense Prediction > Skip connections.
What the sources say
- Not checked — no full text of the sourcecites [38] 38. Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, and Hervé Jégou. Training data-efficient image transformers & distillation through at- tention. arXiv preprint arXiv:2012.12877, 2020. source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“Note that the DeIT-Base architecture is identical to ViT-Base, while DeIT-Base-Dist introduces an additional distillation token, which we ignore in the Reassemble operation.”
§ Vision Transformers for Dense Prediction > Skip connections.
What the sources say
- Not checked — no full text of the sourcecites [38] 38. Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, and Hervé Jégou. Training data-efficient image transformers & distillation through at- tention. arXiv preprint arXiv:2012.12877, 2020. source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Similar to prior work on transformers, DPT unfolds its full potential when trained on large-scale datasets.”
§ Vision Transformers for Dense Prediction > 5. Conclusion
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“Note that results for DPT and Mi-DaS are computed at the same input resolution (384 pixels).”
§ Vision Transformers for Dense Prediction > Additional information
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
The paper relies on this without support.
The choice of backbone architecture has a large influence on the capabilities of the overall model, because information lost in the encoder cannot be recovered in the decoder.
“However, it is widely recognized that the choice of backbone architecture has a large influence on the capabilities of the overall model, as any information that is lost in the encoder is impossible to recover in the decoder.”
The paper relies on this without support.
Feature resolution and granularity are critical for dense prediction, where the architecture should ideally resolve features at or close to the input resolution.
“While feature resolution and granularity may not matter for some tasks, such as image classification, they are critical for dense prediction, where the architecture should ideally be able to resolve features at or close to the resolution of the input image.”
The paper relies on this without support.
Convolutions are linear operators with a limited receptive field.
“Convolutions, by definition, are linear operators that have a limited receptive field.”
The paper relies on this without support.
Downsampling intermediate representations is necessary to keep memory consumption at levels feasible with existing computer architectures.
“Downsampling the intermediate representations is necessary to keep memory consumption at levels that are feasible with existing computer architectures.”
The paper relies on this without support.
The transformer backbone maintains a constant feature resolution.
“Since the transformer backbone maintains a constant feature resolution, it is not clear at which points in the backbone features should be tapped.”
The paper relies on this without support.
Performance of fully-convolutional architectures suffers heavily when performing inference at an input resolution significantly different from the training resolution.
“Performance thus suffers heavily when performing inference at an input resolution that is significantly different from the training resolution.”
The paper relies on this without support.
Transformer encoders have a global receptive field in every layer.
“Transformer encoders, on the other hand, have a global receptive field in every layer.”
The paper relies on this without support.
The reduced meta-dataset datasets provide high-quality ground truth.
“We choose these datasets since they provide high-quality ground truth.”
The paper relies on this without support.
Convolutional architectures offer natural points of interest for passing features from the encoder to the decoder, namely before or after downsampling of the representation.
“Convolutional architectures offer natural points of interest for passing features from the encoder to the decoder, namely before or after downsampling of the representation.”
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“Vision transformers, which are a deep learning approach to encoding low-dimensional input into a high dimensional feature space, have established new frontiers in depth estimation compared to convolutional neural networks (Ranftl et al., 2021).”
From Very high resolution canopy height maps from RGB imagery using self-supervised vision transformer and convolutional decoder trained on aerial lidar (2024) · cites this paper as Ranftl et al., 2021
In the citing paper
“However, the Transformer framework does not involve dimensionality reduction10,12, thus keeping all attention layer traceable to the original input features13, thus making the models interpretable.”
From Transformer for one stop interpretable cell type annotation (2023) · cites this paper as [10]
In the citing paper
“For example, the change of dimensions and non-linear aggregation of features throughout the autoencoder’s deep processing stages leads to untraceable and uninterpretable latent space and loss of information and feature resolution9,10.”
From Transformer for one stop interpretable cell type annotation (2023) · cites this paper as [10]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Vision Transformers for Dense Prediction > 4.1. Monocular Depth Estimation
“Our architecture matches or improves state-of-the-art performance on both datasets in all metrics.”
§ Vision Transformers for Dense Prediction > 5. Conclusion
“| BTS [20] | 0.956 | 0.993 | 0.998 | 0.059 2.756 | 0.096 | | DPT-Hybrid 0.959 | 0.995 | 0.999 | 0.062 2.573 | 0.092 |”
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 4×
“We perform a similar experiment with the hybrid architecture in Table 6 (bottom), where R0 and R1 refer to us-ing features from the first and second downsampling stages of the ResNet50 embedding network.”
MethodInfluentialVaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404
U-Net: Convolutional Networks for Biomedical Image Segmentation
Ronneberger, O., Fischer, P., & Brox, T. (2015) · Lecture notes in computer science · doi:10.1007/978-3-319-24574-4_28 · Cited by 99,404 · cited 3×
“These include training at higher input resolution (if the computational budget permits), dilated convolutions [51] to rapidly increase the receptive field without downsampling, appropriately-placed skip connections from multiple stages of the encoder to the decoder [33], or, more recently, by connecting multiresolutio…”
BackgroundAdam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173
ImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333 · cited 5×
“The encoder is frequently based on an image classification network, also called the backbone, that is pretrained on a large corpus such as ImageNet [9].”
InfluentialBERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325 · cited 6×
“Following work in NLP, the ViT additionally adds a special token that is not grounded in the input image and serves as the final, global image representation which is used for classification.”
BackgroundMethodInfluentialAn Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al. (2020) · arXiv · doi:10.48550/arxiv.2010.11929 · Cited by 47,472 · cited 8×
“On a high level, the vision transformer (ViT) [11] operates on a bag-of-words representation of the image [38].”
BackgroundMethodResultInfluentialFully convolutional networks for semantic segmentation
Long, J., Shelhamer, E., & Darrell, T. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298965 · Cited by 38,834 · cited 2×
“Virtually all existing architectures for dense prediction are based on convolutional networks [6, 33, 36, 44, 51, 52, 55].”
BackgroundLanguage Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372
Research works printed but not linked: 3
- yuan, 2020, Objectcontextual representations for semantic segmentation, ECCV
- godard, 2017, Unsupervised monocular depth estimation with leftright consistency, CVPR
- casser, 2019, Unsupervised learning of depth and ego-motion: A structured approach, AAAI
Cited by
Sorted by citations
Items 1,201–1,300 of 2,248
Incorporating Learned Depth Perception Into Monocular Visual Odometry to Improve Scale Recovery
Mailka, H., Abouzahir, M., & Ramzi, M. (2023) · International Journal of Advanced Computer Science and Applications · doi:10.14569/ijacsa.2023.01408115 · Cited by 2
“In contrast to the state-of-the-art CNN-based method, Ran-ftl et al. [61] reported improved relative performance using the dense prediction transformer (DPT) model for monocular depth estimation.”
BackgroundMethodInfluentialDiffRoad: A Conditional Diffusion-Based Network for Accurate Road Extraction
Yu, Z., Chen, Z., Chen, Y., Hu, R., He, Q., Sun, L., Chen, X., Sun, Z., Xiao, K., & Guo, H. (2026) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2026.3687930 · Cited by 2
Geometric-Aware Low-Light Image and Video Enhancement via Depth Guidance
Lin, Y., Xu, X., Wu, J., Han, Y., & Liu, Z. (2025) · IEEE Transactions on Image Processing · doi:10.1109/tip.2025.3597046 · Cited by 5
“By training on extensive datasets that encompass diverse scenes and objects, current depth estimation models (e.g., DPT [13]) exhibit remarkable zero-shot performance on various images and videos.”
BackgroundMethodZhang, Y., Cheung, M., Yang, C., Zhai, X., Shen, Z., Ji, X., Fu, E. Y., Chau, S. Y., & Luo, D. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5375557 · Cited by 2
Song, Z., Zhu, R., Wang, J., Wang, C., He, J., Deng, J., Yang, W., & Zhang, T. (2025) · ACM Transactions on Multimedia Computing Communications and Applications · doi:10.1145/3750050 · Cited by 3
“…under challenging conditions, while our method can robustly deal with perturbations, delivering accurate and consistent depth predictions in both domains. depth acquisition through 3D sensors (e.g., LiDAR), estimating depth from a single image has garnered widespread attention due to its…”
Out-of-distribution monocular depth estimation with local invariant regression
Hu, Y., Rao, Y., Yu, H. L., Wang, G., Fan, H., Pang, W., & Dong, J. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.113518 · Cited by 2
Bao, N., Carballo, A., Tsukada, M., & Takeda, K. (2023) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc57777.2023.10422148 · Cited by 2
“In recent years, Vision Transformers (ViTs) [15], [16], [17] have emerged as a powerful approach for visual recognition and understanding tasks.”
BackgroundHuBot: A biomimicking mobile robot for non-disruptive bird behavior study
Saoud, L. S., Lesobre, L., Sorato, E., Qaydi, S. A., Hingrat, Y., Seneviratne, L. D., & Hussain, I. (2024) · Ecological Informatics · doi:10.1016/j.ecoinf.2024.102939 · Cited by 4
Bring Your Own Grasp Generator: Leveraging Robot Grasp Generation for Prosthetic Grasping
Stracquadanio, G., Vasile, F., Maiettini, E., Boccardo, N., & Natale, L. (2025) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra55743.2025.11127995 · Cited by 2
“We rely on a model with a DPT [25] head and a DINOv2 [9] backbone for monocular depth estimation.”
Physics-Informed Knowledge Transfer for Underwater Monocular Depth Estimation
Yang, J., Gong, M., & Pu, Y. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-73209-6_26 · Cited by 4
Niu, Y., Li, X., & Wang, Y. (2025) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2025.3627221 · Cited by 2
“Since some of our baselines need depth from depth estimation as prior information, we use DPT [52] to generate the monocular depth as input.”
Optical metasurfaces for general vision processing on the edge
Peng, J., Luo, M., Han, Y., Wu, S., Li, H., Shastri, B. J., Shu, C., Dou, Q., Chai, Y., & Huang, C. (2026) · Nature · doi:10.1038/s41586-026-10635-z · Cited by 6
Monocular Depth Estimation Primed by Salient Point Detection and Normalized Hessian Loss
Huynh, L., Pedone, M., Nguyen-Ha, P., Matas, J., Rahtu, E., & Heikkilä, J. (2021) · International Conference on 3D Vision (3DV) · doi:10.1109/3dv53792.2021.00033 · Cited by 2
“In case of NYU-Depth-v2, the best approaches including TransDepth [65], SARPN [3], DPT [48], VNL [66], BTS [34], and DAV [25] use 38.4, 25.9, 15.3, 14.1, 5.8 and 3.1 times more parameters in contrast to ours, respectively.”
BackgroundMethodInfluentialSelf-attention Guidance Based Crowd Localization and Counting
Ma, Z., Gu, G., & Zhao, W. (2024) · Machine Intelligence Research · doi:10.1007/s11633-023-1428-6 · Cited by 2
4D-VGGT: A SpatioTemporal Foundation Model for Dynamic Scene Geometry Estimation
Wang, H., Zhou, H., Liu, H., & Yan, L. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37035-8_7 · Cited by 2
ZRG: A Dataset for Multimodal 3D Residential Rooftop Understanding
Corley, I., Lwowski, J., & Najafirad, P. (2024) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv57701.2024.00457 · Cited by 4
“We use and modify the the segmentation model architectures described in Section 5.2 by fixing the output layer to contain a single channel with a ReLU activation [12] for continuous outputs, similarly to the architectures used in [38].”
MethodInfluentialControlling Geometric Abstraction and Texture for Artistic Images
Büßemeyer, M., Reimann, M., Buchheim, B., et al. (2023) · International Conference on Cyberworlds (CW) · doi:10.1109/cw58918.2023.00011 · Cited by 2
“For automated stylization, we test saliency prediction (using LPF [32]) and depth prediction (using DPT [33]) to interpolate between two parameter sets (Fig.”
MethodOcclusion-Aware Amodal Depth Estimation for Enhancing 3D Reconstruction From a Single Image
Jo, S.-U., Lee, D. Y., & Rhee, C. E. (2024) · IEEE Access · doi:10.1109/access.2024.3436570 · Cited by 3
“DPT [32] effectively enhances relative depth performance outcomes using a vision transformer (ViT) [34] encoder structure and token fusion through reassembly.”
BackgroundMethodLandgraf, S., Qin, R., & Ulrich, M. (2025) · Qeios · doi:10.32388/yljos8 · Cited by 2
SMFR-Net: simple multi-domain flare removal network
Liu, S., Chen, G., Zhang, W., et al. (2025) · Scientific Reports · doi:10.1038/s41598-025-21378-8 · Cited by 2
3D Photo Stylization: Learning to Generate Stylized Novel Views from a Single ImagePreprint
Mu, F., Wang, J., Wu, Y., et al. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2112.00169 · Cited by 2
“Strong depth estimators such as DPT [45] and LeReS [60] fail on image style transfer outputs from AdaIN [20], LST [31], and AdaAttN [35], because stylized images do not follow natural image statistics.”
BackgroundSmallgs: Gaussian Splatting-Based Camera Pose Estimation for Small-Baseline Videos
Yao, Y., Zhang, Y., Huang, Z., et al. (2025) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw67362.2025.00437 · Cited by 2
“In CF-3DGS, the sparse points used for initialization are replaced by lifting the predicted depth using an off-the-shelf monocular depth estimation network, DPT [19].”
Buffer Anytime: Zero-Shot Video Depth and Normal from Image Priors
Kuang, Z., Zhang, T., Zhang, K., Tan, H., Bi, S., Hu, Y., Xu, Z., Hašan, M., Wetzstein, G., & Luan, F. (2025) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52734.2025.01645 · Cited by 2
“Depth Anything V2 [55] is a Dense Vision Transformer (DPT) [41] that consists of a Vision Transformer (ViT) [13] as an encoder, and a lightweight refinement network that fuses the feature outputs of several ViT blocks together and produces the final results.”
Dong, L., Ren, Q., Shi, J., et al. (2025) · IEEE Transactions on Consumer Electronics · doi:10.1109/tce.2025.3542973 · Cited by 2
Self-Supervised Monocular Depth Estimation Using HOG Feature Prediction
He, X., & Zhao, X. (2024) · International Conference on Computer and Multimedia Technology (ICCMT) · doi:10.1145/3675249.3675316 · Cited by 2
Jing, Y., & Zhang, L. (2025) · doi:10.1109/eect64505.2025.10966991 · Cited by 2
“Ranftl [21] proposed a vision Transformer-based model for intensive prediction tasks, enhancing comparison robustness in complex scenes.”
Robust Initialization and High-Accuracy State Estimation for Filtering-Based Visual-Inertial System
Dong, R., Le, X., Quan, M., et al. (2023) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3298420 · Cited by 2
“Benefiting from deep learning-based monocular depth estimation [16], [17], [18] used the learned monocular depth as input to improve the robustness of visual inertial odometry (VIO) initialization in scenarios with small parallax and low motion excitation.”
MethodImage depth estimation assisted by multi-view projection
Liu, L., Tian, J., Luo, G., Xu, S., Zhang, C., Hu, H., & Tao, W. (2024) · Complex & Intelligent Systems · doi:10.1007/s40747-024-01688-6 · Cited by 2
“We compare our depth estimation method with several classic depth estimation algorithms [7, 8, 23, 50], as well as the top-performing methods on the dataset leaderboard: NBTR-Net, DAHT [51], and CMANet.”
Kolmogorov–Arnold Network for Remote Sensing Image Semantic Segmentation
Wang, Z., Hu, Y., Zhang, X., et al. (2026) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2026.3713199 · Cited by 2
“In contrast, transformer-based models have been introduced into the field of computer vision [19], [20], [21], leveraging multihead self-attention (MSA) to capture global contextual information and model long-range dependencies.”
CoProU-VO: Combining Projected Uncertainty for End-to-End Unsupervised Monocular Visual Odometry
Xie, J., Dhaouadi, O., Chen, W., Meier, J., Kaiser, J., & Cremers, D. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-12840-9_32 · Cited by 2
DualMatch: A Dual EMA Teacher for Underwater Semi-Supervised Pipeline Segmentation
Silva, E. L., Schein, T. T., Brião, S. L., et al. (2025) · SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI) · doi:10.1109/sibgrapi67909.2025.11223277 · Cited by 2
AquaFormer: Color degradation aware transformer for underwater image enhancement
Prakash, K. N., & Jasmine, K. P. (2025) · Signal Image and Video Processing · doi:10.1007/s11760-025-04265-7 · Cited by 2
“These AUVs are increasingly utilized for several critical tasks, including object detection [1] and depth estimation [2], all of which rely on high-quality visual data.”
Mahdavi, A., Haghighi, M. M., & Khankalantary, S. (2024) · doi:10.1109/icee63041.2024.10668382 · Cited by 2
“In this study, Five recent deep learning-based monocular depth estimation methods are compared: GPLDepth [17], Adabins [18], 3DShape [19], Midas [20], and DPT [21].”
MethodMONO-HYDRA: REAL-TIME 3D SCENE GRAPH CONSTRUCTION FROM MONOCULAR CAMERA INPUT WITH IMU
Udugama, U. V. B. L., Vosselman, G., & Nex, F. (2023) · ISPRS annals of the photogrammetry, remote sensing and spatial information sciences · doi:10.5194/isprs-annals-x-1-w1-2023-439-2023 · Cited by 2
Xie, Y., Li, X., Wang, D., Wang, Z., Li, S., Wang, J., & Sun, Z. (2026) · Sensors · doi:10.3390/s26072096 · Cited by 2
Li, S., Bai, H., Gao, Y., Lv, C., Yuan, H., Li, C., Hua, W., & Xie, T. (2025) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2025.3613146 · Cited by 4
“M ONOCULAR depth estimation (MDE) [1], [2], [3], [4], [5], [6], [7] is to estimate the 3D structure of a scene from a single image, which is represented as a depth map.”
Stereo Vision Conversion from Planar Videos Based on Temporal Multiplane Images
Diao, S., Chen, Y., Zhao, Y., et al. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i2.27917 · Cited by 2
“In each mask branch, the depth map D t of frame I t is firstly obtained by a monocular depth estimation model DPT (Ranftl, Bochkovskiy, and Koltun 2021).”
BackgroundMethodPre-trained Self-Attention Framework: An Efficient Mechanism for Source Separation
Tan, H. M., Fhadli, M., Quoc, N. K., et al. (2024) · Lecture notes in networks and systems · doi:10.1007/978-3-031-74127-2_9 · Cited by 2
Liu, Y., & Yan, T. (2025) · Image and Vision Computing · doi:10.1016/j.imavis.2025.105633 · Cited by 2
Redefining Accuracy: Underwater Depth Estimation for Irregular Illumination Scenes
Liu, T., Zhang, S., & Yu, Z. (2024) · Sensors · doi:10.3390/s24134353 · Cited by 2
“To further demonstrate the effectiveness of integrating auxiliary depth information into the self-supervised network, we compared our model with depth estimation meth-ods based on supervised large models [25,50].”
MethodTransformer-based 3D range-gated imaging method with multiple depth priors
Liu, X., Zhang, D., Gao, J., & Wang, X. (2025) · Optics & Laser Technology · doi:10.1016/j.optlastec.2025.113234 · Cited by 2
Enhancing Monocular Depth Estimation with Multi-Source Auxiliary Tasks
Quercia, A., Yildiz, E., Cao, Z., Krajsek, K., Morrison, A., Assent, I., & Scharr, H. (2025) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv61041.2025.00627 · Cited by 2
“In particular DPT [51] is the first attempt to Figure 1.”
InfluentialAdversarial Patch Detection: Leveraging Depth Contrast for Enhanced Threat Visibility
Bunzel, N., & Hamborg, J. (2024) · International Conference on Dependable Systems and Networks workshops · doi:10.1109/dsn-w60302.2024.00020 · Cited by 2
“More recent approaches, such as DPT [31] and DPT-DinoV2 [32] utilizing vision transformers. c) Adversarial Patch Attacks: An adversarial patch attack introduces a perturbation to a specific region of an image with the intention of misleading image classifiers [12] or manipulating the results of…”
BackgroundMethodFusion of Explainable Deep Learning Features Using Fuzzy Integral in Computer Vision
Wang, Y., Pedrycz, W., Ishibuchi, H., & Zhu, J. (2024) · IEEE Transactions on Fuzzy Systems · doi:10.1109/tfuzz.2024.3434589 · Cited by 2
Distribution bias embedding tuning of vision transformer for remote sensing object detection
Liang, Y., Wang, Y., Xie, X., Wang, K., Wang, Y., Zhang, H., Li, Z., Zhou, L., Zhang, Z., & Shi, Y. (2025) · Scientific Reports · doi:10.1038/s41598-025-29376-6 · Cited by 2
“Alexey et al 5 . deployed a dense prediction head for the ViT to pursue fine-grained predictions and reached a new state of the art on several semantic segmentation datasets.”
Liu, X., Wu, B., Wei, C., & Yan, X. (2024) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2024.3409450 · Cited by 2
“These variants have achieved state-of-the-art results in tasks such as segmentation [71], object detection [72], and dense prediction [73].”
BackgroundGraSS: Graph Neural Networks for Loop Closure Detection with Semantic and Spatial Assistance
Lu, S., Peng, Z., Xiao, Z.-L., Yan, B., Lin, S.-S., Yu, S., & He, D. (2024) · International Symposium on Circuits and Systems · doi:10.1109/iscas58744.2024.10557928 · Cited by 2
“Depths d i of keypoints are extracted using DPT [30].”
MethodMohadikar, P., & Duan, Y. (2025) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv61041.2025.00783 · Cited by 2
“For example, Ren ´ e Ranftl [33] proposed using a vision transformer for the MDE task, while CMT [19] used CNN and a Trans-former to capture local and global information that improves feature aggregation.”
ViT-MPI: Vision Transformer Multiplane Images for Surgical Single-View View Synthesis
Han, C., Shao, R., Wu, G., Shao, H., & Liu, Y. (2024) · Lecture notes in computer science · doi:10.1007/978-981-99-8850-1_3 · Cited by 2
ColonAdapter: Geometry Estimation Through Foundation Model Adaptation for Colonoscopy
Jiang, Z., Wang, Y., Cheng, X., & Ge, Z. (2025) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2025.3623938 · Cited by 2
“Following [27], deeper transformer layers capture high-level semantics, while shallow layers retain low-level details.”
ReShader: View-Dependent Highlights for Single Image View-Synthesis
Paliwal, A., Nguyen, B. G., Tsarov, A., & Kalantari, N. K. (2023) · ACM Transactions on Graphics · doi:10.1145/3618393 · Cited by 2
“The three channel novel camera position vector is converted to a 125-channel feature vector using a multilayer perceptron (MLP) with a series of fully connected layers.”
MethodDriving Safe Speed Estimation Based on Outside Environment Vision Analysis
Kashevnik, A. M., & Ali, A. (2022) · Conference of Open Innovations Association (FRUCT) · doi:10.23919/fruct56874.2022.9953827 · Cited by 2
Sparse Transformer-based bins and Polarized Cross Attention decoder for monocular depth estimation
Wang, H., Du, J., Song, K., & Cui, L. (2024) · Engineering Science and Technology an International Journal · doi:10.1016/j.jestch.2024.101705 · Cited by 2
HumanRAM: Feed-forward Human Reconstruction and Animation Model using Transformers
Yu, Z., Li, Z., Bao, H., Yang, C., & Zhou, X. (2025) · ACM SIGGRAPH Conference (SIGGRAPH) · doi:10.1145/3721238.3730605 · Cited by 2
“Inspired by the dense prediction transformers (DPT) used in various vision transformer models [Oquab et al. 2023; Ranftl et al. 2021; Wang et al. 2024; Yang et al. 2024b], we replace the linear layer with stacks of residual CNN layers, similar to DPT heads, enhancing the local information fusion.”
Hyperspherical Learning in Multi-Label Classification
Ke, B., Zhu, Y., Li, M., et al. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19806-9_3 · Cited by 2
“Recently, vision transformer has gained popularity in the field of image recognition [11][40][31][37][3].”
BackgroundHafeez, M. A., Sistu, G., Madden, M. G., & Ullah, I. (2025) · Signal Image and Video Processing · doi:10.1007/s11760-025-04496-8 · Cited by 2
“The performance of an MDE model depends upon several factors: (i) architecture [10]; (ii) loss function [12]; and (iii) the quantity and quality of the training data [13].”
InfluentialFA-MSVNet: multi-scale and multi-view feature aggregation methods for stereo 3D reconstruction
Li, Y., Zhou, Y. Z., Zhao, J., Du, W.-L., & Yao, R. (2024) · Multimedia Tools and Applications · doi:10.1007/s11042-024-20431-4 · Cited by 2
“In recent years, attention mechanisms [6] have demonstrated their advantages in capturing global contextual information, leading to significant improvements in tasks such as object detection, image classification, and semantic segmentation [7, 8].”
Monocular Depth Estimation for Autonomous Driving Based on Instance Clustering Guidance
Kim, D., Jin, D., & Kim, C. (2024) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.1109/apsipaasc63619.2025.10849076 · Cited by 2
“Although depth features are produced through the information exchange between pixels and clusters in (1) and (2), those processes, in themselves, do not process object instances and background regions separately.”
Song, C., Peng, J., Chen, Q., et al. (2025) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2025.108856 · Cited by 2
Depth-Guided NeRF Training via Earth Mover’s Distance
Rau, A., Aklilu, J. G., Holsinger, F. C., & Yeung, S. Y. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-73039-9_1 · Cited by 3
Liu, Y., Zhang, Q., Chen, X., Xu, C., Wang, Q., & Liu, J. (2024) · Physics of Fluids · doi:10.1063/5.0221881 · Cited by 2
MBUDepthNet: Real-Time Unsupervised Monocular Depth Estimation Method for Outdoor Scenes
Bian, Z., Wang, X., Liu, Q., Lv, S., & Wei, R. (2024) · IEEE Access · doi:10.1109/access.2024.3396084 · Cited by 2
“By applying the same method, Ranftl et al. replaced all convolutional layers in their neural network and built a vision transformer for dense predictions, which excels most CNNs when predicting depth maps [36].”
MethodEfficient Computation Sharing for Multi-Task Visual Scene Understanding
Shoouri, S., Yang, M., Fan, Z., & Kim, H.-S. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01571 · Cited by 6
“The recent success of transformer models in many downstream vision tasks [36, 12, 77, 60, 48, 37] has led to the emergence of transformer-based multi-task networks [50, 2, 67, 7, 43], aimed at improving multi-task performance even further.”
BackgroundLinStereo: Linear-Complexity Global Attention for Multi-scale Iterative Stereo Matching
Wang, Y., Turner, O., & Ila, V. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37013-6_1 · Cited by 2
RetinaViT: Efficient Visual Backbone for Online Video Streams
Suzuki, T., & Aoki, Y. (2024) · Sensors · doi:10.3390/s24175457 · Cited by 2
“For multi-scale frame-level feature extraction, we upsampled the output of the 4th, 8th, and 12th blocks to different resolutions to fit the BMVOS decoder with deconvolutions, as conducted in DPT [16].”
MethodCollaborative Blind Image Deblurring
Eboli, T., Morel, J., & Facciolo, G. (2024) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw63382.2024.00792 · Cited by 2
“Although depth estimation could be provided by recent monocular depth estimators, e.g., [37], we keep this multimodal approach for future work and focus instead on blurs where the grouping can be done manually grouping can be done manually as for the examples in Figure 3.”
BackgroundLearning Semantic-Level Multi-Modal Alignment Transformer for RGB-D Co-Salient Object Detection
Wu, Y., Hu, S., Zhang, K., et al. (2026) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2026.3668582 · Cited by 2
“We utilize a pre-trained depth estimation model [82] to estimate the depth of RGB images in the DUTS-Class dataset [29] to obtain depth maps.”
Vision Transformer Framework Approach for Yellow Nail Syndrome Disease Identification
Roy, V. K., Thakur, V., & Nijhawan, R. (2022) · Lecture notes in networks and systems · doi:10.1007/978-981-16-7657-4_34 · Cited by 2
Beche, R., & Nedevschi, S. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.02413 · Cited by 2
“This section expands on the design and interpretation of the Delentropic Scene Profile.”
InfluentialZhang, Y., Zhang, G., Li, K., et al. (2025) · PLoS ONE · doi:10.1371/journal.pone.0321878 · Cited by 2
“By optimizing the sampling strategy, enhancing the MLP network, and extracting depth priors from more accurate pre-trained monocular depth models [18] or coarse depth maps generated by consumer-grade depth sensors, we have achieved significant performance gains.”
Robust 3D Shape Reconstruction in Zero-Shot from a Single Image in the Wild
Cho, J., Youwang, K., Yang, H., & Oh, T.-H. (2025) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52734.2025.02122 · Cited by 4
PDDepth: Pose Decoupled Monocular Depth Estimation for Roadside Perception System
Wang, H., Zhang, X., Chen, Z., Li, J., & Liu, H. (2025) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2025.3541313 · Cited by 2
Rapid Network Adaptation:Learning to Adapt Neural Networks Using Test-Time Feedback
Yeo, T., Kar, O. F., Sodagar, Z., & Zamir, A. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00431 · Cited by 5
“Table 4: RNA works across different architectures of the main network fθ such as DPT [80] and ConvNext [61].”
BackgroundMethodInfluential3D Gaussian Splatting for Lunar Surface with Limited View Input
Zhang, X., Cui, L., Zhang, X., et al. (2025) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2025.3619948 · Cited by 2
UDGS-SLAM: UniDepth Assisted Gaussian Splatting for Monocular SLAM
Mansour, M., Abdelsalam, A., Happonen, A., Porras, J., & Rahtu, E. (2025) · Array · doi:10.1016/j.array.2025.100400 · Cited by 3
“Focusing on MMDE [31,34–41], it aims to predict absolute values in physical units (e.g., meters), which is necessary to perform 3D reconstruction effectively.”
Zaytoon, M., & Torki, M. (2023) · Proceedings - IEEE Symposium on Computers and Communications/IEEE Symposium on Computers and Communications · doi:10.1109/iscc58397.2023.10218197 · Cited by 2
“For the depth estimation module, we rely on the pretrained model provided by [25].”
BackgroundMethodInfluentialTiP4GEN: Text to Immersive Panorama 4D Scene Generation
Xing, K., Liang, H., Xu, D., Yin, Y., Plataniotis, K. N., Zhao, Y., & Wei, Y. (2025) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3746027.3754704 · Cited by 5
Boosting Depth Estimation for Self-Driving in a Self-Supervised Framework via Improved Pose Network
Dayoub, Y., Savchenko, A. V., & Makarov, I. (2024) · IEEE Open Journal of the Computer Society · doi:10.1109/ojcs.2024.3505876 · Cited by 2
“Furthermore, inspired by the success of transformers in computer vision, several studies have explored transformer architectures in the depth estimation networks [31], [56].”
Ciubotariu, G., Tomescu, V.-I., & Czibula, G. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-87156-7_6 · Cited by 2
Bokeh Rendering Based on Adaptive Depth Calibration Network
Liu, L., Zhou, L., & Dong, Y. (2023) · International Conference on Computer and Communication Systems (ICCCS) · doi:10.1109/icccs57501.2023.10151285 · Cited by 2
“This framework integrates the Vision Transformer (VIT) and Vision Transformer for Dense Predictions (DPT) to retain feature maps and obtain a depth map, respectively.”
MethodInfluentialHGM: A General High-Order Spatial and Spectral Global Feature Fusion Module for Visual Multitasking
Chen, C., Zhang, X., Zhou, Y., Chang, Y., & Zeng, W. (2024) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2024.3417414 · Cited by 2
“We were inspired by the second-order spatial information interaction triggered by self-attention [10], [11] and proposed the CS n module [25] to address the issue of CNNs falling short in feature extraction compared with trans-formers.”
BackgroundMethodCoSplat: Resolving multi-view inconsistency in feed-forward 3D Gaussian Splatting
Hu, F., Liu, B., Fang, L., et al. (2026) · Displays · doi:10.1016/j.displa.2026.103549 · Cited by 2
NeRFscopy: Neural Radiance Fields for in-Vivo Time-Varying Tissues from Endoscopy
Salort-Benejam, L., & Agudo, A. (2026) · IEEE International Symposium on Biomedical Imaging (ISBI) · doi:10.1109/isbi61048.2026.11515806 · Cited by 2
Thermopore: Predicting Part Porosity Based on Thermal Images Using Deep LearningPreprint
Pak, P., Ogoke, F., Polonsky, A. T., et al. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4836563 · Cited by 2
Karimi, F., Mehrpanah, A., & Rawassizadeh, R. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4531792 · Cited by 3
Echo Depth Estimation via Attention-based Hierarchical Multi-scale Feature Fusion Network
Zhang, W., Yin, J., Yu, P., et al. (2025) · ACM Transactions on Multimedia Computing Communications and Applications · doi:10.1145/3736768 · Cited by 2
“Ranftl et al. [36], Agarwal and Arora [37], and Zhang et al. [38] proposed leveraging the Transformer as the backbone network.”
2D Human Skeleton Action Recognition Based on Depth Estimation
Wang, L., Shanmin, Y. A. N. G., Zhang, J., et al. (2024) · IEICE Transactions on Information and Systems · doi:10.1587/transinf.2023edp7223 · Cited by 2
“Ranftl et al. [11] used Transformers as feature extractors.”
MethodLo, C.-C., & Vandewalle, P. (2024) · Sensors · doi:10.3390/s24061864 · Cited by 2
“Lo and Vandewalle [34] proposed a dedicated transformer module for radar input, reassembling features from radar and images instead of using readout tokens, as in [26,27].”
BackgroundA Ground Truth-Free Metric for Evaluating Depth Estimation Techniques
Zaitoon, A. (2023) · International Mobile, Intelligent, and Ubiquitous Computing Conference (MIUCC) · doi:10.1109/miucc58832.2023.10278314 · Cited by 2
“When large-scale training data is available, DPT [3] achieves an outstanding performance in general-purpose monocular depth estimation[20], also the model architecture is able to be fine-tuned on monocular depth estimation datasets, like KITTI [10].”
BackgroundMethodBridging Spectral-Wise and Multi-Spectral Depth Estimation Via Geometry-Guided Contrastive Learning
Shin, U., Lee, K., & Oh, J. (2025) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra55743.2025.11127755 · Cited by 4
“The networks utilize convolution layer [7], [8], [12] and trans-former block [10], [13]–[15] to project input image into non-linear feature space.”
Improvement of Self-supervised Depth and Motion Learning with Vision Transformer
Ma, X.-Q., Nguyen, N.-K., & Nguyen, V.-T. (2022) · Lecture notes in networks and systems · doi:10.1007/978-981-19-3394-3_13 · Cited by 2
Enhancing Self‐Supervised Monocular Depth Estimation in Endoscopy via Feature‐Based Perceptual Loss
Zhu, K., & Cui, L. (2025) · IET Image Processing · doi:10.1049/ipr2.70035 · Cited by 2
Giffard, E., Jannin, P., & Baxter, J. S. H. (2023) · International Journal of Computer Assisted Radiology and Surgery · doi:10.1007/s11548-023-03023-9 · Cited by 2
“For image segmentation, more modern architectures such as those derived from U-Nets [22] or vision transformers [23, 24] have distinctly top-down and bottom-up components working in tandem, demonstrating that the computer vision community is moving away from the purely bottom-up approaches it…”
Eye-for-an-eye: Appearance Transfer with Dense Semantic Correspondence in Diffusion Models
Go, S., Choi, K., Shin, M., & Uh, Y. (2026) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv61042.2026.00451 · Cited by 2
3D Reconstruction of Endoscopic Images Using Depth Anything Model v2
Kim, T. D., & Bui, D.-M. (2025) · Lecture notes in networks and systems · doi:10.1007/978-3-032-00267-9_7 · Cited by 2
Ulreich, F., Kaup, A., & Ebert, M. (2024) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc58415.2024.10920009 · Cited by 3
Enhancing monocular depth estimation with an advanced encoder-decoder architecture
El-Alfy, Y. E., Baroudi, U. A., & Luqman, H. (2025) · Neural Computing and Applications · doi:10.1007/s00521-025-11566-y · Cited by 2
“Supervised-based Learning Ranftl et al. [15] proposed a dense vision architecture for depth estimation.”
Jing, D., Mo, H., Han, L., et al. (2025) · International Journal of Communication Systems · doi:10.1002/dac.70109 · Cited by 2
Zhao, X., Li, X., Liu, T., & Shen, G. (2024) · Energy · doi:10.1016/j.energy.2024.132866 · Cited by 2
Metadata
- Authors
- Ranftl, R., Bochkovskiy, A., & Koltun, V.
- Affiliations
- Intel Labs
- Intel (United States) (US)
- Venue
- 2021 IEEE/CVF International Conference on Computer Vision (ICCV)
- Publisher
- IEEE
- Published
- 2021-10
- Pages
- 12159-12168
- OpenAlex
- W3136635488
- Cited by
- 2,373
- References
- 88
- Citations by year
- 2026: 480 · 2025: 607 · 2024: 586 · 2023: 431 · 2022: 225 · 2021: 44
- FWCI
- 104.94
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- TIB Data Manager
- IEEE/CVF International Conference on Computer Vision (ICCV)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Vision and Imaging
- Keywords
- vision transformer, dense prediction, monocular depth estimation, semantic segmentation, convolutional decoder, global receptive field, Pascal Context, NYUv2, KITTI, ADE20K
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-10-11, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-11 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
