Xception: Deep Learning with Depthwise Separable Convolutions
The DOI matches its Crossref record; 20 of the 21 references it cites were found in a registry or an index.
Abstract
We present an interpretation of Inception modules in convolutional neural networks as being an intermediate step in-between regular convolution and the depthwise separable convolution operation (a depthwise convolution followed by a pointwise convolution). In this light, a depthwise separable convolution can be understood as an Inception module with a maximally large number of towers. This observation leads us to propose a novel deep convolutional neural network architecture inspired by Inception, where Inception modules have been replaced with depthwise separable convolutions. We show that this architecture, dubbed Xception, slightly outperforms Inception V3 on the ImageNet dataset (which Inception V3 was designed for), and significantly outperforms Inception V3 on a larger image classification dataset comprising 350 million images and 17,000 classes. Since the Xception architecture has the same number of parameters as Inception V3, the performance gains are not due to increased capacity but rather to a more efficient use of model parameters.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
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?
In the citing paper
“The top-5 popular baselines adopted in the evaluation are XceptionNet (Chollet 2017), Afchar et al. (2018), Nguyen et al. (2019a), ResNet (He et al. 2016), and Yang et al. (2019b).”
From Countering Malicious DeepFakes: Survey, Battleground, and Horizon (2022) · cites this paper as Chollet 2017
In the citing paper
“The main idea behind Xception is its depthwise separable convolution [125].”
From Review of deep learning: concepts, CNN architectures, challenges, applications, future directions (2021) · cites this paper as [125]
In the citing paper
“Depthwise separable convolution (DSC) has been found in past research to have a smaller number of parameters than traditional convolution [26].”
From Multimodal Emotion Recognition Based on Facial Expressions, Speech, and EEG (2024) · cites this paper as [26]
In the citing paper
“In the research [24], a depthwise separable convolution neural network on FPGA is being reviewed due to its ability to reduce the number of model parameters and improve calculation speeds, which makes it widely used in mobile edge computing applications such as Xception [25] and MobileNet [26].”
From Embedded Intelligence on FPGA: Survey, Applications and Challenges (2021) · cites this paper as [25]
In the citing paper
“For classification (Table 10), the performance of Xception [12], LipForensics [21], Face X-Ray [36], and SBI [56] also drops compared to LAV-DF.”
From AV-Deepfake1M: A Large-Scale LLM-Driven Audio-Visual Deepfake Dataset (2024) · cites this paper as [12]
5 more claims about this paper
Create a free account to see every claim other papers make about this one.
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 2×
“in [4], which our proposed architecture uses extensively.”
BackgroundVery Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 2×
“In 2012, these ideas were refined into the AlexNet architecture [9], where convolution operations were being repeated multiple times in-between max-pooling operations, allowing the network to learn richer features at every spatial scale.”
BackgroundGoing deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562
ImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380 · cited 2×
“Since its first introduction, Inception has been one of the best performing family of models on the ImageNet dataset [14], as well as internal datasets in use at Google, in particular JFT [5].”
BackgroundMethodRethinking the Inception Architecture for Computer Vision
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.308 · Cited by 32,138
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Ioffe, S., & Szegedy, C. (2015) · arXiv · doi:10.48550/arxiv.1502.03167 · Cited by 27,818
MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., & Adam, H. (2017) · arXiv · doi:10.48550/arxiv.1704.04861 · Cited by 19,015
Visualizing and understanding convolutional networks
Zeiler, M. D., & Fergus, R. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-10590-1_53 · Cited by 15,839 · cited 2×
“What followed was a trend to make this style of network increasingly deeper, mostly driven by the yearly ILSVRC competition; first with Zeiler and Fergus in 2013 [25] and then with the VGG architecture in 2014 [18].”
BackgroundDistilling the Knowledge in a Neural Network
Hinton, G., Vinyals, O., & Dean, J. (2015) · arXiv · doi:10.48550/arxiv.1503.02531 · Cited by 14,204 · cited 2×
“Since its first introduction, Inception has been one of the best performing family of models on the ImageNet dataset [14], as well as internal datasets in use at Google, in particular JFT [5].”
Background
Research works printed but not linked: 7
- lin, 2013, Network in Network
- sifre, 2014, Rigid-motion scattering for image classification
- silberman, 2016, Tf-slim
- chollet, 2015, Keras
- abadi, 2015, Tensor-Flow Large-scale machine learning on heterogeneous systems
- tieleman, 0, Divide the gradient by a running average of its recent magnitude, COURSERA Neural Networks for Machine Learning, 4
- vanhoucke, 2014, Learning visual representations at scale, ICLRE
Cited by
Items 1,101–1,200 of 1,227
Li, G., Jiang, S., Yun, I., Kim, J., & Kim, J. (2020) · IEEE Access · doi:10.1109/access.2020.2971760 · Cited by 53
Bilal, M., Maqsood, M., Yasmin, S., Hasan, N. U., & Rho, S. (2022) · The Journal of Supercomputing · doi:10.1007/s11227-021-03957-4 · Cited by 51
“eXtreme Inception (Xception) [38], inspired by original Inception architecture is a deep convolutional neural network with 71 layers.”
BackgroundBayoudh, K., Hamdaoui, F., & Mtibaa, A. (2020) · Physical and Engineering Sciences in Medicine · doi:10.1007/s13246-020-00957-1 · Cited by 50
A Review of Deepfake Techniques: Architecture, Detection and Datasets
Edwards, P., Nebel, J.-C., Greenhill, D., & Liang, X. (2024) · IEEE Access · doi:10.1109/access.2024.3477257 · Cited by 50
“In 2017, Chollet [33] presented the Xception network as a parameter-efficient adaptation to the Inception-V3 [34] architecture, where stackable depthwise separable convolu-tions are used as a replacement for the Inception module while achieving notably improved performance.”
MethodImproving Generalization of Deepfake Detection With Data Farming and Few-Shot Learning
Korshunov, P., & Marcel, S. (2022) · IEEE Transactions on Biometrics, Behavior, and Identity Science · doi:10.1109/tbiom.2022.3143404 · Cited by 50
“For that reason, we have selected two state of the art algorithms, based on Xception model [18] and EfficientNet variant B4 [2], both of which showed a great performance on several databases [1] and were also successfully used in many top systems from Deepfake Detection Challenge [7].”
MethodKhan, S. A., & Dang-Nguyen, D.-T. (2024) · IEEE Access · doi:10.1109/access.2023.3348450 · Cited by 53
DeepFake-Adapter: Dual-Level Adapter for DeepFake Detection
Shao, R., Wu, T., Nie, L., & Liu, Z. (2025) · International Journal of Computer Vision · doi:10.1007/s11263-024-02274-6 · Cited by 48
“…we compare the linear separability regarding real/fake faces of FaceForensics++ dataset (Rossler et al., 2019) based on features extracted by Xception (Chollet, 2017) pre-trained on ImageNet-1K (Deng et al., 2009) and ViT-Base (Dosovit-skiy et al., 2020) pre-trained on ImageNet-21K, respectively.”
InfluentialFully Unsupervised Deepfake Video Detection Via Enhanced Contrastive Learning
Qiao, T., Xie, S., Chen, Y., Retraint, F., & Luo, X. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3356814 · Cited by 48
“Thus, it is proposed to discuss the baseline backbone networks [53], [54], [55], [56], [57], [58] to verify the validity of our selection for Deepfake video detection.”
BackgroundDeepfake Detection with Deep Learning: Convolutional Neural Networks versus Transformers
Thing, V. L. L. (2023) · 2023 IEEE International Conference on Cyber Security and Resilience (CSR) · doi:10.1109/csr57506.2023.10225004 · Cited by 48
Deepfake detection: humans vs. machines
Korshunov, P., & Marcel, S. (2020) · arXiv · doi:10.48550/arxiv.2009.03155 · Cited by 47
GBDF: Gender Balanced DeepFake Dataset Towards Fair DeepFake Detection
Nadimpalli, A. V., & Rattani, A. (2023) · Lecture Notes in Computer Science · doi:10.1007/978-3-031-37742-6_25 · Cited by 46
On the Generalization of Deep Learning Models in Video Deepfake Detection
Coccomini, D. A., Caldelli, R., Falchi, F., & Gennaro, C. (2023) · Preprint (MDPI) · doi:10.20944/preprints202303.0161.v1 · Cited by 46
“…X-ray [42] FF++ 65.5 Patch-based [43] FF++ 65.6 DSP-FWA [44] FF++ 67.3 CSN [45] FF++ 68.1 Multi-Task [46] FF++ 68.1 CNN-GRU [47] FF++ 68.9 Xception [48] FF++ 70.9 CNN-aug [49] FF++ 72.1 LipForensics [50] FF++ 73.5 FTCN [10] FF++ 74.0 RealForensics [45] FF++ 75.9 RealForensics [45] FF++ 75.9…”
InfluentialMobile Health (mHealth) Viral Diagnostics Enabled with Adaptive Adversarial Learning
Shokr, A., Pacheco, L. G. C., Thirumalaraju, P., Kanakasabapathy, M. K., Gandhi, J., Kartik, D., Silva, F. S. R., Erdogmus, E., Kandula, H., Luo, S., et al. (2021) · ACS Nano · doi:10.1021/acsnano.0c06807 · Cited by 45
“30 For HBV, HIV, and ZIKV, the source data were loaded in batches by random sampling.”
MethodDeep Learning based Framework for Automatic Damage Detection in Aircraft Engine Borescope Inspection
Shen, Z., Wan, X., Ye, F., Guan, X., & Liu, S. (2019) · 2019 International Conference on Computing, Networking and Communications (ICNC) · doi:10.1109/iccnc.2019.8685593 · Cited by 45
“We implement a similar algorithm which uses a deep learning architecture based on Xception [24] and train it on our labeled data with transfer learning.”
MethodA Collaborative Federated Learning Framework for Lung and Colon Cancer Classifications
Hossain, M. M., Islam, M. R., Ahamed, M. F., Ahsan, M., & Haider, J. (2024) · Technologies · doi:10.3390/technologies12090151 · Cited by 44
“The Inception-V3, VGG16, ResNet50, ResNeXt50, and Xception models achieved average classification accuracies of 98.96%, 98.36%, 98.88%, 98.88%, and 99.10%, respectively The Inception-V3, VGG16, ResNet50, ResNeXt50, and Xception models achieved average classi fi cation accuracies of 98.96%, 98.36%,…”
InfluentialDetection and Localization of Facial Expression Manipulations
Mazaheri, G., & Roy-Chowdhury, A. K. (2022) · 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) · doi:10.1109/wacv51458.2022.00283 · Cited by 44
Exploring Adversarial Fake Images on Face Manifold
Li, D., Wang, W., Fan, H., & Dong, J. (2021) · 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr46437.2021.00573 · Cited by 43
Al-Timemy, A. H., Khushaba, R. N., Mosa, Z. M., & Escudero, J. (2021) · Studies in Systems, Decision and Control · doi:10.1007/978-3-030-69744-0_6 · Cited by 41
“v2 Sandler et al. [32] 224x224x3 Logits 154 3.5 10 ShuffleNet Zhang et al. [48] 224x224x3 node_202 172 1.4 11 GoogLeNet Szegedy et al. [49] 224x224x3 loss3-classifier 144 7 12 Xception Chollet et al. [50] 299x299x3 predictions 170 22.9 13 NASNet-Mobile Zoph et al. [51] 224x224x3 predictions 913 5.3 14 NASNet-Large Zop…”
InfluentialFace Forgery Detection by 3D Decomposition and Composition Search
Zhu, X., Fei, H., Zhang, B., Zhang, T., Zhang, X., Li, S. Z., & Lei, Z. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2022.3233586 · Cited by 41
“In the CS, the encoding operation E is the first seven blocks of Xception [63] adjust the channel size of the backbone to 1 = 8 to reduce the computation dependence.”
MethodForgery-Aware Adaptive Learning With Vision Transformer for Generalized Face Forgery Detection
Luo, A., Cai, R., Kong, C., Ju, Y., Kang, X., Huang, J., & Kot, A. C. (2025) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2024.3522091 · Cited by 41
Yurtsever, E., Liu, Y., Lambert, J., Miyajima, C., Takeuchi, E., Takeda, K., & Hansen, J. H. L. (2019) · 2019 IEEE Intelligent Transportation Systems Conference (ITSC) · doi:10.1109/itsc.2019.8917362 · Cited by 40
“761 FT+LSTM Xception [38] ImageNET 23 50 0.”
MethodBurgos-Artizzu, X. P., Coronado-Gutiérrez, D., Valenzuela-Alcaraz, B., Vellvé, K., Eixarch, E., Crispi, F., Bonet-Carne, E., Bennasar, M., & Gratacos, E. (2021) · American Journal of Obstetrics & Gynecology MFM · doi:10.1016/j.ajogmf.2021.100462 · Cited by 39
Deep learning-based classification of microalgae using light and scanning electron microscopy images
Sonmez, M. E., Altinsoy, B., Ozturk, B. Y., Gumus, N. E., & Eczacioglu, N. (2023) · Micron · doi:10.1016/j.micron.2023.103506 · Cited by 38
Borra, D., Fantozzi, S., & Magosso, E. (2021) · Frontiers in Human Neuroscience · doi:10.3389/fnhum.2021.655840 · Cited by 36
“This design, besides performing temporal convolution in the first layer, uses separable and depthwise convolutions, i.e., convolutions specifically devoted to reducing the number of trainable parameters (Chollet, 2016).”
BackgroundBeyond Deepfake Images: Detecting AI-Generated Videos
Vahdati, D. S., Nguyen, T. D., Azizpour, A., & Stamm, M. C. (2024) · 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw63382.2024.00443 · Cited by 37
Explainable AI for the Choquet Integral
Murray, B. J., Islam, M. A., Pinar, A. J., Anderson, D. T., Scott, G. J., Havens, T. C., & Keller, J. M. (2021) · IEEE Transactions on Emerging Topics in Computational Intelligence · doi:10.1109/tetci.2020.3005682 · Cited by 36
“Herein, we fuse seven DCNNs — CaffeNet [40], GoogleNet [41], ResNet 50 [42], ResNet 101, DenseNet [43], InceptionResNetV2 [44], and Xception [45] — on the AID remote sensing data set [46].”
MethodKaur, S., Kumar, P., & Kumaraguru, P. (2020) · Journal of Electronic Imaging · doi:10.1117/1.jei.29.3.033013 · Cited by 35
MINTIME: Multi-Identity Size-Invariant Video Deepfake Detection
Coccomini, D. A., Zilos, G. K., Amato, G., Caldelli, R., Falchi, F., Papadopoulos, S., & Gennaro, C. (2024) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2024.3409054 · Cited by 33
“Consider M I i ∈ { 0 , 1 } ( NHW +1) × 1 as the mask of identity I i indicating to the model which faces attend to an identity and which to another during the attention calculation.”
BackgroundMethodEmergence of deepfakes and video tampering detection approaches: A survey
Kingra, S., Aggarwal, N., & Kaur, N. (2023) · Multimedia Tools and Applications · doi:10.1007/s11042-022-13100-x · Cited by 33
“Triplet Network using XceptionNet, RNN and 3-”
MethodInfluentialLip Sync Matters: A Novel Multimodal Forgery Detector
Shahzad, S. A., Hashmi, A., Khan, S., Peng, Y.-T., Tsao, Y., & Wang, H.-M. (2022) · 2022 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.23919/apsipaasc55919.2022.9980296 · Cited by 33
“Additionally, fake video detection is empowered by many unimodal video/image methods, such as Capsule Foren-sics [21], HeadPose [22], Xception [23], LipForensics [24], Meso-4 and MesoInception4 [25].”
MethodMFF-Net: Deepfake Detection Network Based on Multi-Feature Fusion
Zhao, L., Zhang, M., Ding, H., & Cui, X. (2021) · Entropy · doi:10.3390/e23121692 · Cited by 32
“[41] 87.00% 82.00% 74.00% 74.00% CustomPooling CNN [42] 80.00% 62.00% 59.00% 59.00% MesoNet [10] 90.00% 83.00% 83.00% 75.00% Xception [36] 96.01% 93.29% 94.71% 79.14% Slowfast [48] 97.53% 94.93% 95.01% 82.55% SPSL [45] 93.48% 86.02% 92.26% 76.78% F 3 -net(Xception) [19] 97.97% 95.32% 96.53% 83.32%…”
BackgroundInfluentialAppearance Matters, So Does Audio: Revealing the Hidden Face via Cross-Modality Transfer
Kong, C., Chen, B., Yang, W., Li, H., Chen, P., & Wang, S. (2022) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2021.3057457 · Cited by 32
Multimodal Forgery Detection Using Ensemble Learning
Hashmi, A., Shahzad, S. A., Ahmad, W., Lin, C. W., Tsao, Y., & Wang, H.-M. (2022) · 2022 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.23919/apsipaasc55919.2022.9980255 · Cited by 32
“Moreover, [11] and [18], respectively, proposed highlevel and mesoscopic features, [12] adopted a capsule network, [37] designed a XceptionNet, and [38] derived a two-stream”
BackgroundAttack Agnostic Dataset: Towards Generalization and Stabilization of Audio DeepFake Detection
Kawa, P., Plata, M., & Syga, P. (2022) · Interspeech 2022 · doi:10.21437/interspeech.2022-10078 · Cited by 33
“During tests, we used the following solutions: LCNN [10], MesoInception– 4 [27], XceptionNet [28], RawNet2 [12] and GMM [9].”
MethodInfluentialSch-net: a deep learning architecture for automatic detection of schizophrenia
Fu, J., Yang, S., He, F., He, L., Li, Y., Zhang, J., & Xiong, X. (2021) · BioMedical Engineering OnLine · doi:10.1186/s12938-021-00915-2 · Cited by 31
“DenseNet and Xception gain slightly higher accuracies than ResNet, owing to the networks contains both the residual module and effective algorithms to reduce the number of parameters.”
BackgroundMethodInfluentialAn improved EEGNet for single-trial EEG classification in rapid serial visual presentation task
Zhang, H., Wang, Z., Yu, Y., Yin, H., Chen, C., & Wang, H. (2022) · Brain Science Advances · doi:10.26599/bsa.2022.9050007 · Cited by 30
“It includes the depthwise convolution and pointwise convolution layers [29] with parameter F2.”
BackgroundIF-CNN: Image-Aware Inference Framework for CNN With the Collaboration of Mobile Devices and Cloud
Shu, G., Liu, W., Zheng, X., & Li, J. (2018) · IEEE Access · doi:10.1109/access.2018.2880196 · Cited by 30
Multimodal Approach for DeepFake Detection
Lomnitz, M., Hampel-Arias, Z., Sandesara, V., & Hu, S. (2020) · 2020 IEEE Applied Imagery Pattern Recognition Workshop (AIPR) · doi:10.1109/aipr50011.2020.9425192 · Cited by 30
Acoustic Features Analysis for Explainable Machine Learning-Based Audio Spoofing Detection
Bisogni, C., Loia, V., Nappi, M., & Pero, C. (2024) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2024.104145 · Cited by 30
An adversarial attack approach for eXplainable AI evaluation on deepfake detection models
Gowrisankar, B., & Thing, V. L. L. (2024) · Computers & Security · doi:10.1016/j.cose.2023.103684 · Cited by 30
Subject adaptation convolutional neural network for EEG-based motor imagery classification
Liu, S., Zhang, J., Wang, A., Wu, H., Zhao, Q., & Long, J. (2022) · Journal of Neural Engineering · doi:10.1088/1741-2552/ac9c94 · Cited by 29
“(c) Feature Fusion: Given the features generated by the depthwise convolution layer, we stack and fit them into a pointwise convolution layer [49] with a kernel size of (1× 1), and each convolution kernel can fuse the input features into the corresponding vector.”
Detection of perineural invasion in prostate needle biopsies with deep neural networks
Kartasalo, K., Ström, P., Ruusuvuori, P., Samaratunga, H., Delahunt, B., Tsuzuki, T., Eklund, M., & Egevad, L. (2022) · Virchows Archiv · doi:10.1007/s00428-022-03326-3 · Cited by 29
“Convolutional DNNs were used to classify patches (Xception) and to identify the regions in the biopsy where PNI was present (Unet) [15, 16].”
MethodAbbasi, M., Váz, P., Silva, J., & Martins, P. (2025) · Applied Sciences · doi:10.3390/app15031225 · Cited by 32
Zhang, F., Yang, F., Li, C., & Yuan, G. (2019) · IEEE Access · doi:10.1109/access.2019.2919103 · Cited by 29
“MIO-TCD is used to evaluate classification algorithms (i.e., AlexNet [11], InceptionV3 [34], ResNet50 [14], VGG-19 [12], Xception [35], and DenseNet [36]) and detection algorithms (i.e., Faster R-CNN [17], SSD-300, SSD-512 [22], YOLO [21] and YOLOv2 [23]), and achieves superior performance.”
InfluentialA Comparative Study: Deepfake Detection Using Deep-learning
Khatri, N., Borar, V., & Garg, R. (2023) · 2023 13th International Conference on Cloud Computing, Data Science & Engineering (Confluence) · doi:10.1109/confluence56041.2023.10048888 · Cited by 29
“The variance in accuracy is least for XceptionNet, followed by InceptionV3, VGG16 and MobileNetV2.”
MethodInfluentialSeidel, R., Jahn, N., Seo, S., Goerttler, T., & Obermayer, K. (2022) · IEEE Open Journal of Intelligent Transportation Systems · doi:10.1109/ojits.2021.3139393 · Cited by 28
Automatic detection of airway invasion from videofluoroscopy via deep learning technology
Lee, S. J., Ko, J. Y., Kim, H. I., & Choi, S.-I. (2020) · Applied Sciences · doi:10.3390/app10186179 · Cited by 28
“Therefore, among the models that can be effectively trained on limited data, in this paper, we choose a CNN based on an Xception module [27].”
MethodDeepfake Detection by Exploiting Surface Anomalies: The Surfake Approach
Ciamarra, A., Caldelli, R., Becattini, F., Seidenari, L., & Del Bimbo, A. (2024) · 2024 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW) · doi:10.1109/wacvw60836.2024.00112 · Cited by 27
“Although we are processing patches that describe surfaces of faces rather than canonical RGB face images, a significant amount of test samples have been correctly separated in the projection space.”
MethodDiagonalwise Refactorization: An Efficient Training Method for Depthwise Convolutions
Qin, Z., Zhang, Z., Li, D., Zhang, Y., & Peng, Y. (2018) · 2018 International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.2018.8489312 · Cited by 27
Suratkar, S., Johnson, E., Variyambat, K., Panchal, M., & Kazi, F. (2020) · 2020 11th International Conference on Computing, Communication and Networking Technologies (ICCCNT) · doi:10.1109/icccnt49239.2020.9225400 · Cited by 27
“Finally, the XceptionNet [18] which implements depthwise separable convolutions, while using the InceptionNet as its backbone, was used.”
MethodNarayan, K., Agarwal, H., Thakral, K., Mittal, S., Vatsa, M., & Singh, R. (2022) · 2022 IEEE International Joint Conference on Biometrics (IJCB) · doi:10.1109/ijcb54206.2022.10007968 · Cited by 26
“For this, we propose DeePhy, a Deepfake Phylogeny dataset and benchmark the existing state-of-the-art techniques that are popular in deepfake literature [33, 34, 35, 36, 37].”
BackgroundMethodEnhancing General Face Forgery Detection via Vision Transformer with Low-Rank Adaptation
Kong, C., Li, H., & Wang, S. (2023) · 2023 IEEE 6th International Conference on Multimedia Information Processing and Retrieval (MIPR) · doi:10.1109/mipr59079.2023.00033 · Cited by 33
“…4000 fake videos generated by four manipulation techniques: Deepfake(DF), [14] 0.741 0.673 0.648 0.600 0.634 0.594 0.598 0.567 0.655 0.609 Xception [8] 0.766 0.694 0.696 0.643 0.626 0.593 0.597 0.552 0.671 0.621 EfficientNet [38] 0.451 0.485 0.537 0.505 0.512 0.503 0.499 0.497 0.500 0.498 All-train…”
InfluentialA hierarchical feature selection strategy for deepfake video detection
Mohiuddin, S., Sheikh, K. H., Malakar, S., Velásquez, J. D., & Sarkar, R. (2023) · Neural Computing and Applications · doi:10.1007/s00521-023-08201-z · Cited by 25
“The proposed method uses fine-tuned XceptionNet [7] network to learn the important features from face images.”
BackgroundMethodInfluentialWeakly-supervised deepfake localization in diffusion-generated images
Ţânţaru, D.-C., Oneaţă, E., & Oneaţă, D. (2024) · 2024 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) · doi:10.1109/wacv57701.2024.00614 · Cited by 25
“Then we detail the generative techniques that we are interested in detecting (§3.2) and the datasets generated with these methods (§3.3).”
MethodInfluentialPrediction of protein subcellular localization in single cells
(2025) · doi:10.1038/s41592-025-02696-1 · Cited by 25
“To adapt the ESM-2 representation to predict protein localization, a light attention layer 17 using separable convolutions 16 , which was found to perform well in protein sequence to localization predictions 13 , is applied to the ESM-2 representation to obtain a 300-dimensional sequence…”
InfluentialExposing DeepFakes Using Convolutional Neural Networks and Transfer Learning Approaches
Suratkar, S., Kazi, F., Sakhalkar, M., Abhyankar, N., & Kshirsagar, M. (2020) · 2020 IEEE 17th India Council International Conference (INDICON) · doi:10.1109/indicon49873.2020.9342252 · Cited by 25
“Xception has performed better than Inception V3 in the ImageNet challenge [37].”
MethodVisual-Semantic Transformer for Face Forgery Detection
Xu, Y., Jia, G., Huang, H., Duan, J., & He, R. (2021) · 2021 IEEE International Joint Conference on Biometrics (IJCB) · doi:10.1109/ijcb52358.2021.9484407 · Cited by 24
Park, J., Kwak, I.-Y., & Lim, C. (2021) · Electronics · doi:10.3390/electronics10161996 · Cited by 24
“Khan et al. [57] proposed CoroNet using Xception architecture pretrained on ImageNet dataset.”
BackgroundA Joint Model for Multimodal Document Quality Assessment
Shen, A., Salehi, B., Baldwin, T., & Qi, J. (2019) · 2019 ACM/IEEE Joint Conference on Digital Libraries (JCDL) · doi:10.1109/jcdl.2019.00024 · Cited by 24
“A wide range of models have been proposed to tackle the image classification task, such as VGG (Simonyan and Zisser-man 2014), ResNet (He et al. 2016), Inception V3 (Szegedy et al. 2016), and Xception (Chollet 2017).”
Pancreatic Adenocarcinoma: Imaging Modalities and the Role of Artificial Intelligence
Anghel, C., Mugur, G. C., Anghel, D. A., Rusu-Munteanu, G. I., Dumitru, R. L., & Lupescu, I. G. (2024) · doi:10.20944/preprints202401.0904.v1 · Cited by 23
Contrastive Learning for DeepFake Classification and Localization via Multi-Label Ranking
Hong, C.-Y., Hsu, Y.-C., & Liu, T.-L. (2024) · 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52733.2024.01669 · Cited by 23
Swin-Fake: A Consistency Learning Transformer-Based Deepfake Video Detector
Gong, L. Y., Li, X. J., & Chong, P. H. J. (2024) · Electronics · doi:10.3390/electronics13153045 · Cited by 22
Videoforensicshq: Detecting High-Quality Manipulated Face Videos
Fox, G., Liu, W., Kim, H., Seidel, H.-P., Elgharib, M., & Theobalt, C. (2021) · 2021 IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme51207.2021.9428101 · Cited by 22
“As one example, FaceForensics++ [8] contains internet videos modified by several face synthesis techniques [1, 3, 11–13] and demonstrates that an off-the-shelf image classifier, XceptionNet [14], outperforms methods specifically designed for fake detection.”
MethodInfluentialDimitri, G. M. (2022) · Computers · doi:10.3390/computers11110163 · Cited by 22
Hidden Path Selection Network for Semantic Segmentation of Remote Sensing Images
Yang, K., Tong, X.-Y., Xia, G.-S., Shen, W., & Zhang, L. (2022) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2022.3197334 · Cited by 22
Vision Transformers: From Semantic Segmentation to Dense Prediction
Zhang, L., Lu, J., Zheng, S., Zhao, X., Zhu, X., Fu, Y., Xiang, T., Feng, J., & Torr, P. H. S. (2024) · International Journal of Computer Vision · doi:10.1007/s11263-024-02173-w · Cited by 23
“Window embedding operations can be implemented by depth-wise convolution [56], average pooling or max pooling with kernel size and stride equal to the window size.”
InfluentialZhang, J., Petitjean, C., & Ainouz, S. (2022) · Journal of Imaging · doi:10.3390/jimaging8020023 · Cited by 21
DeSI: Deepfake Source Identifier for Social Media
Narayan, K., Agarwal, H., Mittal, S., Thakral, K., Kundu, S., Vatsa, M., & Singh, R. (2022) · 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw56347.2022.00323 · Cited by 21
“Deepfake Detection: We employ a deep learning network namely XceptionNet [4] for deepfake detection.”
MethodInfluentialSpatio-Temporal FAST 3D Convolutions for Human Action Recognition
Stergiou, A., & Poppe, R. (2019) · 2019 18th IEEE International Conference On Machine Learning And Applications (ICMLA) · doi:10.1109/icmla.2019.00036 · Cited by 21
“, [5], [10], [13], [14], [31]), pre-training on larger datasets (e.”
Jahanifar, M., Tajeddin, N. Z., Koohbanani, N. A., Gooya, A., & Rajpoot, N. (2018) · arXiv · doi:10.48550/arxiv.1809.10243 · Cited by 21
“The pre-trained networks that we utilized in this framework are 152-layered ResNet (ResNet152) [18], 169-layered DensNet (DensNet169) [19], Xception [20], and Inception-ResNet v2 (ResNetV2) [21].”
MethodPalmer, M. S., Huebner, S. E., Willi, M., Fortson, L., & Packer, C. (2021) · Human Computation · doi:10.15346/hc.v8i2.123 · Cited by 20
AOT: Appearance Optimal Transport Based Identity Swapping for Forgery Detection
Zhu, H., Fu, C., Wu, Q., Wu, W., Qian, C., & He, R. (2020) · arXiv · doi:10.48550/arxiv.2011.02674 · Cited by 20
Huang, J., Kuchaiev, O., O'Neill, P., Lavrukhin, V., Li, J., Flores, A., Kucsko, G., & Ginsburg, B. (2020) · arXiv · doi:10.48550/arxiv.2005.04290 · Cited by 20
Large-Scale Whale Call Classification Using Deep Convolutional Neural Network Architectures
Wang, D., Zhang, L., Lu, Z., & Xu, K. (2018) · 2018 IEEE International Conference on Signal Processing, Communications and Computing (ICSPCC) · doi:10.1109/icspcc.2018.8567758 · Cited by 19
“As the newest model in the Inception model group, Xception architecture [13] employs the depth-wise separable convolution operation to replace the regular Inception modules, which has become a cornerstone of convolutional neural network architecture design.”
MethodLi, S., Liu, F., Peng, G., Cheng, F., Zhao, B., & Ji, M. (2022) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2022.3214300 · Cited by 18
“For Xception on the QUGS dataset, the Flops and Param are 2.82 and 3.56 times larger with 6.54% lower result.”
BackgroundMethodInfluentialDetecting CNN-Generated Facial Images in Real-World Scenarios
Hulzebosch, N., Ibrahimi, S., & Worring, M. (2020) · arXiv · doi:10.48550/arxiv.2005.05632 · Cited by 18
“Based on the work of [13, 28, 42], we select Xception [7] and ForensicTransfer [13] as state-of-the-art model architectures for CNN-generated image detection.”
BackgroundMethodInfluentialSketch2code: Generating a website from a paper mockup
Robinson, A. (2019) · arXiv · doi:10.48550/arxiv.1905.13750 · Cited by 18
“We fine tuned an Xception model [87] (a CNN based on Inception v3 [88]) pretrained on ImageNet [89] - a large real world image database - to use as a backbone network for Deeplab v3+ [56].”
Method(2020) · doi:10.1109/cvprw50498.2020.00329 · Cited by 18
“Based on the work of [13, 28, 42], we select Xception [7] and ForensicTransfer [13] as state-of-the-art model architectures for CNN-generated image detection.”
BackgroundMethodInfluentialDeep learning for Amazon satellite image analysis
Bragilevsky, L., & Bajic, I. V. (2017) · 2017 IEEE Pacific Rim Conference on Communications, Computers and Signal Processing (PACRIM) · doi:10.1109/pacrim.2017.8121895 · Cited by 17
“In addition to the custom model, we used several other well-known CNN architectures such as Xception [14], VGG16/19 [15], ResNet50 [16], and InceptionV3 [17].”
MethodImproving Facial Emotion Recognition Systems with Crucial Feature Extractors
Pandey, R. K., Karmakar, S., Ramakrishnan, A. G., & Saha, N. (2019) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-30642-7_24 · Cited by 17
A defensive attention mechanism to detect deepfake content across multiple modalities
Asha, S., Vinod, P., & Menon, V. G. (2024) · Multimedia Systems · doi:10.1007/s00530-023-01248-x · Cited by 16
“Table 1 shows the precision, recall, F1-score, and accuracy of four CNN architecture models: VGG16 [43], EfficientNet [50], InceptionResNetV2 [51], and XceptionNet [52].”
On the Vulnerability of Deepfake Detectors to Attacks Generated by Denoising Diffusion Models
Ivanovska, M., & Štruc, V. (2024) · 2024 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW) · doi:10.1109/wacvw60836.2024.00115 · Cited by 16
“The vulnerability of self–supervised methods, is evaluated using 3 models, i.e. DSP-FWA [19], Face X-Ray [18], and Self Blended Images (SBI) [30].”
BackgroundMethodTariq, S., Woo, S. S., Singh, P., Irmalasari, I., Gupta, S., & Gupta, D. (2025) · Proceedings of the 33rd ACM International Conference on Multimedia · doi:10.1145/3746027.3755786 · Cited by 15
"Better Be Computer or I'm Dumb": A Large-Scale Evaluation of Humans as Audio Deepfake Detectors
Warren, K., Tucker, T., Crowder, A., Olszewski, D., Lu, A., Fedele, C., Pasternak, M., Layton, S., Butler, K., Gates, C., et al. (2024) · Proceedings of the 2024 on ACM SIGSAC Conference on Computer and Communications Security · doi:10.1145/3658644.3670325 · Cited by 15
RaScaNet: Learning Tiny Models by Raster-Scanning Images
Yoo, J., Lee, D., Son, C., Jung, S., Yoo, B., Choi, C., Han, J.-J., & Han, B. (2021) · 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr46437.2021.01346 · Cited by 14
“Xception [6], MobileNet [15, 16, 31], and EfficientNet [33] are well-known models that use depthwise separable convolutions.”
BackgroundMethodHelping Visually Impaired People Take Better Quality Pictures
Mandal, M., Ghadiyaram, D., Gurari, D., & Bovik, A. C. (2023) · IEEE Transactions on Image Processing · doi:10.1109/tip.2023.3282067 · Cited by 14
“P2P++ performed the best on average, whereas, the models with Xception [54] and ResNet-50V2 [53] backbones also performed very well.”
MethodResultInfluentialSemi-Supervised Deep Domain Adaptation for Deepfake Detection
Seraj, M. S., Singh, A., & Chakraborty, S. (2024) · 2024 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW) · doi:10.1109/wacvw60836.2024.00116 · Cited by 14
“Following [31], we used the Xception [15] as the back-bone deep neural network architecture in our experiments.”
MethodSkip-attention encoder–decoder framework for human motion prediction
Zhang, R., Shu, X., Yan, R., Zhang, J., & Song, Y. (2022) · Multimedia Systems · doi:10.1007/s00530-021-00807-4 · Cited by 13
A Novel Deep Learning Model for Sea State Classification Using Visual-Range Sea Images
Umair, M., Hashmani, M. A., Hussain Rizvi, S. S., Taib, H., Abdullah, M. N., & Memon, M. M. (2022) · Symmetry · doi:10.3390/sym14071487 · Cited by 13
“Xception The Xception model is inspired by inception architecture [45].”
MethodKim, Y., Kim, H.-I., Park, G. S., Kim, S. Y., Choi, S.-I., & Lee, S. J. (2021) · Journal of Clinical Medicine · doi:10.3390/jcm10122681 · Cited by 12
“Within the ROI, the presence of laryngeal penetration or aspiration was classified by the deep learning network trained with the Xception module [24].”
MethodGANprintR: Improved Fakes and Evaluation of the State of the Art in Face Manipulation Detection
Neves, J. C., Tolosana, R., Vera-Rodriguez, R., Lopes, V., Proenca, H., & Fierrez, J. (2020) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2020.3007250 · Cited by 12
Pham, N., Jia, H., Tran, M., Dinh, T., Bui, N., Kwon, Y., Ma, D., Nguyen, P., Mascolo, C., & Vu, T. (2022) · Proceedings of the 28th Annual International Conference on Mobile Computing And Networking · doi:10.1145/3495243.3560533 · Cited by 12
“Compared to the standard convolution, DW-Conv and PW-Conv require much smaller numbers of parameters, thus being more computationally effective [48].”
MethodSpatio-Temporal Branching for Motion Prediction using Motion Increments
Wang, J., Zhou, Y., Qiang, W., Ba, Y., Su, B., & Wen, J.-R. (2023) · Proceedings of the 31st ACM International Conference on Multimedia · doi:10.1145/3581783.3612330 · Cited by 11
“In contrast, the effectiveness of decoupling spatio-temporal information has been demonstrated in some tasks [5, 47].”
BackgroundHuman Fall Detection with Ultra-Wideband Radar and Adaptive Weighted Fusion
Huang, L., Zhu, A., Qian, M., & An, H. (2024) · Sensors · doi:10.3390/s24165294 · Cited by 11
Yuan, H., Hong, C., Jiang, P.-T., Zhao, G., Tran, N. T. A., Xu, X., Yan, Y. Y., & Liu, N. (2024) · Journal of Biomedical Informatics · doi:10.1016/j.jbi.2024.104673 · Cited by 11
Compressed Deepfake Video Detection Based on 3D Spatiotemporal Trajectories
Chen, Z., Liao, X., Wu, X., & Chen, Y. (2024) · 2024 Asia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.1109/apsipaasc63619.2025.10848944 · Cited by 10
“Landmark localization, a cascade of trained regressors is employed to achieve accurate positioning of facial landmarks in each video frame.”
Sora as a World Model? A Complete Survey on Text-to-Video Generation
Puspitasari, F. D., Zhang, C., Cho, J., Haider, A., Eman, N. U., Amin, O., Mankowski, A., Umair, M., Zheng, J., Zheng, S., et al. (2024) · arXiv · doi:10.48550/arxiv.2403.05131 · Cited by 10
Multisemantic Path Neural Network for Deepfake Detection
Wu, N., Jin, X., Jiang, Q., Wang, P., Zhang, Y., Yao, S., & Zhou, W. (2022) · Security and Communication Networks · doi:10.1155/2022/4976848 · Cited by 9
“For example, the accuracy of Xception [6] was 99.”
InfluentialMitra, A., Chakravarty, A., Ghosh, N., Sarkar, T., Sethuraman, R., & Sheet, D. (2020) · 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC) · doi:10.1109/embc44109.2020.9175246 · Cited by 9
A Novel Cognitively Inspired Deep Learning Approach to Detect Drivable Areas for Self-driving Cars
Jiang, F., Wang, Z., & Yue, G. (2024) · Cognitive Computation · doi:10.1007/s12559-023-10215-7 · Cited by 9
“We compare our method with different backbones, there are MobileNet [46], Xception [47], DRN [48] and ResNet [30].”
Metadata
- Authors
- Chollet, F.
- Affiliations
- Google (United States) (US)
- Corresponding author
- Chollet, Francois
- Venue
- 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
- Publisher
- IEEE
- Published
- 2017-07
- Pages
- 1800-1807
- ISSN
- 1063-6919 (ISSN-L)
- OpenAlex
- W2531409750
- Cited by
- 20,380
- References
- 21
- Citations by year
- 2026: 1618 · 2025: 3098 · 2024: 3287 · 2023: 3281 · 2022: 3003 · 2021: 2735 · 2020: 1859 · 2019: 1096 · 2018: 356 · 2017: 42
- FWCI
- 268.33
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings (publishedVersion)
- 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 › Artificial Intelligence › Machine Learning and Data Classification
- Keywords
- Xception, depthwise separable convolution, Inception module, InceptionV3, convolutional neural network architecture, image classification, parameter efficiency, ImageNet, large-scale image classification
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.
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
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.