Zero-Centered Fixed-Point Quantization With Iterative Retraining for Deep Convolutional Neural Network-Based Object Detectors
Open accessThe DOI matches its Crossref record; 49 of the 63 references it cites were found in a registry or an index, and 1 are not checked yet; 13 of the 27 sources its claims cite were checked for support.
Abstract
In the field of object detection, deep learning has greatly improved accuracy compared to previous algorithms and has been used widely in recent years. However, object detection using deep learning requires many hardware (HW) resources due to the huge computations for high performance, making it very difficult to run real-time on embedded platforms. Therefore, various compression methods have been studied to solve this problem. In particular, quantization methods greatly reduce the computational burden of deep learning by reducing the number of bits used for weights and activation functions in deep learning. However, most of these existing studies targeted only object classification and cannot be applied to object detection. Furthermore, most of the existing quantization studies are based on floating-point operations, which requires additional effort when implementing HW accelerators. This paper proposes an HW-friendly fixed-point-based quantization method that can also be applied to object detection. In the proposed method, the center of the weight distribution is adjusted to zero by subtracting the mean of weight parameters before quantization, and the retraining process is iteratively applied to minimize the accuracy drop caused by quantization. Furthermore, while applying the proposed method to object detection, performance degradation is minimized by considering the minimum and maximum values of weight parameters of deep learning networks. When applying the proposed quantization method to representative one-stage object detectors, You Only Look Once v3 and v4 (YOLOv3 and YOLOv4), detection accuracy similar to the original networks (i.e., YOLOv3 and YOLOv4) with a single-precision floating-point format (32-bit) is maintained despite expressing weights with only about 20% of the bits compared to a single-precision floating-point format in COCO dataset.
Verified 2026-11-19 · 1 reference not checked yet
Overview
CoChat Summary
## TL;DR
The paper proposes an HW-friendly fixed-point quantization method for deep CNNs that combines zero-centering of weights and iterative retraining, and extends it to object detectors with layer-wise bit allocation. On COCO, it reports near-original mAP for YOLOv3 and YOLOv4 while using only about 20% of the original weight bits.
## Key findings
- On CIFAR-10 with VGG-7, applying both zero-centered quantization and iterative retraining reduced accuracy drop to 0.1% at 4/4 bits; the paper reports 91.97% accuracy versus 92.07% for the original network.
- On YOLOv3 with COCO, the proposed method achieved 54.52 mAP at 6.58 average weight bits, compared with 54.64 mAP and 32 bits for the original model; the basic 8-bit quantized model was 52.60 mAP, while the proposed 8-bit model was 54.52 mAP.
- On YOLOv4 with COCO, the proposed method achieved 62.9 mAP at 8 bits versus 64.17 mAP for the original; with bit allocation, it reached 63.54 mAP at 5.99 average weight bits.
- The paper reports that iterative retraining reduced the portion of ambiguous weights from 23.33% to 8.73% in VGG-7, 22.95% to 18.01% in YOLOv3, and 20.06% to 14.65% in YOLOv4.
## Methods
- Proposed a fixed-point quantization scheme that subtracts the mean of layer weights to center them around zero, then quantizes and retrains iteratively while gradually increasing the quantization rate.
- Evaluated on CIFAR-10 with VGG-7 for image classification and COCO with YOLOv3/YOLOv4 for object detection; experiments were run on a Linux server with a GeForce RTX 2080Ti.
- For object detection, added a layer-wise bit allocation strategy using grid search to choose per-layer weight bitwidths that maximize mAP.
## Limitations
- The paper states that basic quantization without the proposed zero-centered and iterative retraining techniques causes severe performance degradation on object detection, even at 8 bits.
- The text makes evident that the method still has noticeable mAP loss at lower bitwidths for object detection: for example, YOLOv3 drops to 50.05 mAP at 6 bits, 46.95 at 5 bits, and 42.02 at 4 bits even with the proposed method.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures

















Verification
Claims the paper makes
5 Valid · 3 Not valid · 11 Not checked · 60 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Most recent quantization studies [21] [23] have focused on INT quantization with integer parameters and oatingpoint scale factors to prevent an accuracy decrease in networks that perform image classi cation.”
What the sources say
- Partly supportedcites [21] [21] S. Jung, C. Son, S. Lee, J. Son, J. J. Han, Y . Kwak, S. J. Hwang, and C. Choi, ''Learning to quantize deep networks by optimizing quantization intervals with task loss,'' in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR) , Jun. 2019, pp. 4350 4359. - sourceExperimental — accuracy not yet measured
The Related Work survey (Sec. 2, p. 2) and Sec. 3.1 (p. 3) do support part of the claim: recent quantization work is described as fixed-point/low-bit quantization of weights and activations that adds scaling factors (BWN's 'several scaling factors on the binary filters', XNOR-Net's 'filter-wise scaling factor', SYQ's 'pixel-wise or row-wise scaling') and is motivated by avoiding the 'severe accuracy degradation' of low-precision networks on image classification (Introduction, p. 1; Abstract, p. 1). But our copy nowhere states that MOST recent studies focus specifically on integer parameters with FLOATING-POINT scale factors — it never characterises the scale factors as floating-point, and much of what it surveys (BinaryConnect, INQ's logarithmic weights, DoReFa-Net's bitwidth study, Gupta's 16-bit fixed-point training, and its own EXP-Net training strategy) falls outside that description. The sweeping 'most recent studies' framing is therefore an overstatement of what this source reports. (LLM, conf=0.72)
- Could not verifycites [23] [23] S. Han, H. Mao, and W. J. Dally, ''Deep compression: Compressing deep neural networks with pruning, trained quantization and Huffman coding,'' 2015, arXiv:1510.00149 . [Online]. Available: http://arxiv.org/abs/1510.00149 - sourceExperimental — accuracy not yet measured
Our copy of Gaussian YOLOv3 contains no discussion of network quantization: it addresses localization uncertainty in object detection (Sec. 3.1 Gaussian modeling, Sec. 3.3 utilization of uncertainty) and accuracy/speed comparisons on KITTI and BDD (Sec. 4.2, Tables 2-4). Searches for 'quantization / INT quantization / fixed-point integer parameters / floating-point scale factors / image classification' returned only semantic neighbours about detection mAP, FLOPs and Gaussian bbox modeling, with no passage on quantized parameters or scale factors. The claim's point is therefore not addressed by this source, and no alternative work is named in our copy to credit it to. (LLM, conf=0.75)
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.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“The proposed technique can be easily applied to various object detectors including YOLOv3 [10] and YOLOv4 [11].”
What the sources say
- Supportedcites [10] [10] J. Redmon and A. Farhadi, ''YOLOv3: An incremental improvement,'' 2018, arXiv:1804.02767 . [Online]. Available: http://arxiv.org/abs/1804.02767 - source
The citation is attached to YOLOv3 as an example of an object detector. The paper identifies and describes YOLOv3 as a detector in “3. How We Do” (p. 3); the citing authors’ assertion about applying their proposed technique is not attributed to this source. (LLM, conf=0.91)
- Supportedcites [11] [11] A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, ''YOLOv4: Optimal speed and accuracy of object detection,'' 2020, arXiv:2004.10934 . [Online]. Available: http://arxiv.org/abs/2004.10934 - source
The citation is attached to YOLOv3 as an example of an object detector, not to the citing authors’ claim that their proposed technique is easily applicable. Section 2.3 (p. 2) describes YOLOv3 predicting bounding boxes across three scales, confirming that identification. (LLM, conf=0.94)
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 that could be checked couldn’t confirm or dispute this claim.
In the paper
“However, despite using 1 bit more for the activation parameter than the proposed method, the accuracy drop of Miyashita et al. [39] is greater by 0.21% than that of the proposed method.”
What the sources say
- Could not verifycites [39] [39] D. Miyashita, E. H. Lee, and B. Murmann, ''Convolutional neural networks using logarithmic data representation,'' 2016, arXiv:1603.01025 . [Online]. Available: https://arxiv.org/abs/1603.01025 - sourceExperimental — accuracy not yet measured
Our copy of Nayak et al. (2019) contains no comparison with Miyashita et al.: searches for 'Miyashita', for the 0.21% accuracy-drop differential, and for activation-bit comparisons returned nothing. The closest passages are IV.a (p. Table 1), which only says ResNet-18 CIFAR-10 results are shown 'at byte boundary for comparison purposes with published results' without naming any work or giving a differential, and III/IV.b, which describe quantization of model parameters only (convolution kernels, batch-norm and dense layers) with accuracy drops of '~4% up to 6-bit' — no activation bit-widths are reported. Thus neither the 0.21% figure nor the 1-bit activation difference can be checked against this source; nothing in our copy contradicts it either. (LLM, conf=0.68)
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.
- Could not verify
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“In particular, through the comparison result with Gaussian YOLOv3-tiny [13], which enhances the accuracy of a small-scale network, it can be seen that effectively reducing the size of a large-scale network is more ef cient than improving the performance of a smallscale network.”
What the sources say
- Could not verifycites [13] [13] J. Choi, D. Chun, H.-J. Lee, and H. Kim, ''Uncertainty-based object detector for autonomous driving embedded platforms,'' in Proc. 2nd IEEE Int. Conf. Artif. Intell. Circuits Syst. (AICAS) , Aug. 2020, pp. 16 20. - sourceExperimental — accuracy not yet measured
The Methodology and Selection of architecture sections (p. 5) discuss balancing speed, accuracy, and model size, but do not compare shrinking a large network with improving a small one. Searches for Gaussian YOLOv3-tiny and the claimed efficiency comparison found no passage addressing that conclusion. (LLM, conf=0.76)
plainstates a causeSingle case- 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.
- Could not verify
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 results illustrate that the portion of the ambiguous weight parameters is signi cantly reduced during retraining from 23.33%, 22.95%, and 20.06% to 8.73%, 18.01%, and 14.65% in VGG-7, YOLOv3, and YOLOv4, respectively”
plainstates a cause- 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.
30 more claims
Create a free account to see all 79 claims and to verify the 9 references still open.
- It may state two separate claims as one. 37
- The quote only partly states this claim. 4
- Probably not a substantial claim. 11
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
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 1×
“Table 2 illustrates the accuracy results of the existing VGG-7 network ( i.e. , 32-bit full decision) [2] and the quantized VGG-7 networks with the various options in the second column.”
BackgroundImageNet 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 1×
“detection and classification increases dramatically [1]–[4].”
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 1×
“Especially, Jung et al. [21] found that a reduction in accuracy does not occur when using only four bits for both weight and activation parameters, even for the ImageNet dataset [32].”
BackgroundYou Only Look Once: Unified, Real-Time Object Detection
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.91 · Cited by 53,803 · cited 2×
“In other words, YOLOv3 is a fully-convolutional network composed of only 1 × 1 and 3 × 3 size convolution filters and illustrates outstanding accuracy for objects of various sizes while maintaining a high processing speed similar to YOLO [33] and YOLOv2 [34].”
MethodResultMicrosoft COCO: Common Objects in Context
Lin, T.-Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Dollár, P., & Zitnick, C. L. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-10602-1_48 · Cited by 45,460 · cited 1×
“5 illustrates the medians of maximum and minimum weight parameters for each layer when training the VGG-7 network with Cifar-10 dataset [35] and the YOLOv3/v4 networks with COCO dataset [36].”
Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation
Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2014.81 · Cited by 32,684 · cited 1×
“In particular, for DNNbased object detection, classification, and localization are performed simultaneously, which requires vast computation [8]–[11].”
Background(2016) · Cited by 31,590
Learning Multiple Layers of Features from Tiny Images
Krizhevsky, A. (2024) · doi:10.57702/zp44cu3g · Cited by 23,395
SSD: Single Shot MultiBox Detector
Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C.-Y., & Berg, A. C. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46448-0_2 · Cited by 22,128
Cited by
A Survey on Adversarial Attack in the Age of Artificial Intelligence
Kong, Z., Xue, J., Wang, Y., Huang, L., Niu, Z., & Li, F. (2021) · Wireless Communications and Mobile Computing · doi:10.1155/2021/4907754 · Cited by 98
“Then, the neural network model is attacked, and the process is repeated [81].”
Toward AI-Assisted UAV for Human Detection in Search and Rescue Missions
Rizk, M., Slim, F., & Charara, J. (2021) · International Conference on Decision Aid Sciences and Applications (DASA) · doi:10.1109/dasa53625.2021.9682412 · Cited by 58
“1 shows the network architecture of YOLOv3 [16].”
BackgroundKim, N. J., & Kim, H. (2022) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2022.3189496 · Cited by 56
Self-citationA survey of FPGA and ASIC designs for transformer inference acceleration and optimization
Kang, B. J., Lee, H. I., Yoon, S. K., Kim, Y. C., Jeong, S. B., Jeong, S. B., & Kim, H. (2024) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2024.103247 · Cited by 63
HAWQ-V3: Dyadic Neural Network Quantization
Yao, Z., Dong, Z., Zheng, Z., Gholaminejad, A., Yu, J., Tan, E., Wang, L., Huang, Q., Wang, Y., Mahoney, M. W., et al. (2021) · International Conference on Machine Learning · Cited by 281
Vision transformer models for mobile/edge devices: a survey
Lee, S. I., Koo, K., Lee, J. H., Lee, G., Jeong, S., Seongjun, O., & Kim, H. (2024) · Multimedia Systems · doi:10.1007/s00530-024-01312-0 · Cited by 53
“Compact CNN models refer to models designed from the beginning with small model size and low computational complexity without lightweight techniques such as Kim and Kim [27, 28].”
Self-citationApple ripeness identification from digital images using transformers
Xiao, B., Nguyen, M., & Yan, W. Q. (2023) · Multimedia Tools and Applications · doi:10.1007/s11042-023-15938-1 · Cited by 42
“Therefore, crucial deep learning structures such as convolution neural networks (CNN or ConvNet) and recurrent neural networks (RNN) have been boardly employed in natural language processing (NLP) in recent years and have achieved significant progress [17, 18].”
Hong, H., Choi, D., Kim, N.-J., Lee, H., Kang, B., Kang, H., & Kim, H. (2024) · Journal of Real-Time Image Processing · doi:10.1007/s11554-024-01442-8 · Cited by 39
“Quantization is a compression technique that converts 32-bit floating-point data (FP32) into low-precision integer or fixed-point formats [61].”
Self-citationAdversarial AI Testcases for Maritime Autonomous Systems
Walter, M. J., Barrett, A., Walker, D. J., & Tam, K. (2023) · AI Computer Science and Robotics Technology · doi:10.5772/acrt.15 · Cited by 43
“The samples can be iteratively generated by retraining the model and regenerating adversarial training data [36].”
Hong, H., Choi, D., Kim, N. J., & Kim, H. (2024) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2024.3440884 · Cited by 42
“However, the conventional quantization [20] involves two multiplications for dequantization and a division operation along with rounding for quantization, utilizing the floating-point scaling factors obtained from the activations and weights for the n th convolution output ( x n , w n ).”
MethodSelf-citationLiang, L., Zheng, Y., Ge, Q., & Zhang, F. (2022) · Frontiers in Psychology · doi:10.3389/fpsyg.2021.762725 · Cited by 33
“The number of iterations of the CNN model obtained here is basically consistent with the number of iterations obtained by Kim and Kim (2021) in CNN research (Kim and Kim, 2021).”
ResultObject Detection in Autonomous Cyber-Physical Vehicle Platforms: Status and Open Challenges
Balasubramaniam, A., & Pasricha, S. (2023) · Springer eBooks · doi:10.1007/978-3-031-28016-0_17 · Cited by 108
Kim, N. J., & Kim, H. (2023) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2023.3338052 · Cited by 27
“Therefore, network compression and acceleration studies as well as performance improvement studies have been actively conducted for the practical use of CNNs in real-world applications [18], [19], [20], [21], [22].”
BackgroundMethodSelf-citationHAWQV3: Dyadic Neural Network Quantization
Yao, Z., Dong, Z., Zheng, Z., Gholami, A., Yu, J., Tan, E., Wang, L., Huang, Q., Wang, Y., Mahoney, M. W., et al. (2020) · arXiv · doi:10.48550/arxiv.2011.10680
“…with quantization (Asanovic & Morgan, 1991; Bhalgat et al., 2020; Chin et al., 2020; Dong et al., 2019; Hubara et al., 2016; Jacob et al., 2018; Kim & Kim, 2021; Park et al., 2018a; Rastegari et al., 2016; Sharma et al., 2018; Song et al., 2020; Zhang et al., 2018; Zhou et al., 2017a; 2016),…”
Comparative Study of Activation Functions and Their Impact on the YOLOv5 Object Detection Model
Doherty, J., Gardiner, B., Kerr, E., Siddique, N. H., & Manvi, S. S. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-09282-4_4 · Cited by 17
Deep learning-based vehicle detection from orthophoto and spatial accuracy analysis
Bıyık, M. Y., Atik, M. E., & Duran, Z. (2022) · International Journal of Engineering and Geosciences · doi:10.26833/ijeg.1080624 · Cited by 18
Hardware-Friendly Logarithmic Quantization with Mixed-Precision for MobileNetV2
Choi, D., & Kim, H. (2022) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas54282.2022.9869994 · Cited by 14
“The MobileNetV2 weight distribution, in particular, has a narrow range and sharp characteristics [21].”
BackgroundMethodSelf-citationChun, D., Choi, J.-W., Lee, H., & Kim, H. (2023) · IEEE Access · doi:10.1109/access.2023.3280552 · Cited by 18
“In particular, they focus on reducing the latency of each layer and consequently, some studies propose lightweight techniques such as quantization and pruning, which can reduce the network size but result in a loss of accuracy [24], [25], [26], [27].”
BackgroundSelf-citationGirdhar, N., Raj, A., Sharma, D., Singh, V. K., Doucet, A., & Renz, M. (2025) · Soft Computing · doi:10.1007/s00500-025-10854-y · Cited by 23
Meteor detection and localization using YOLOv3 and YOLOv4
Alowais, A., Sharif, M. E., Ghali, S., et al. (2023) · Neural Computing and Applications · doi:10.1007/s00521-023-08575-0 · Cited by 10
“3 YOLOv4 Architecture [21] set to 100,000 with momentum of 0.9 and decay of 0.005.”
BackgroundInfluentialA lightweight and efficient one-stage detection framework
Huang, J., Chen, J., & Wang, H. (2022) · Computers & Electrical Engineering · doi:10.1016/j.compeleceng.2022.108520 · Cited by 11
Choi, D., Park, J., & Kim, H. (2024) · IEEE Access · doi:10.1109/access.2024.3488093 · Cited by 13
“Therefore, various studies are being actively conducted on the development of compressed networks such as MobileNetV2 [5] and lightweight techniques such as quantization [10], [11], [12], [13], [14], [15] and pruning [16], [17].”
BackgroundSelf-citationVLCQ: Post-training quantization for deep neural networks using variable length coding
Abdel‐Salam, R., Abdel-Gawad, A. H., & Wassal, A. G. (2024) · Future Generation Computer Systems · doi:10.1016/j.future.2024.107654 · Cited by 10
Accurate Medicine Recognition and Audio Conversion for Handwritten Prescriptions
Edlabadkar, S. R., & Joshi, P. P. (2024) · International Conference on Emerging Smart Computing and Informatics (ESCI) · doi:10.1109/esci59607.2024.10497304 · Cited by 6
“The architecture of YOLOv4 [19]–[21] or You Only Look Once version 4, as shown in the figure 2, is designed to achieve optimal speed and accuracy in object detection tasks.”
ACC: Adaptive Compression Framework for Efficient On-device CNN Training
Lee, G., Lee, S., & Kim, H. (2024) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas59952.2024.10595863 · Cited by 7
“Quantization has numerous advantages, including reducing the bit-width of the CNN parameters, solving the lack of memory resources, and improving energy efficiency by enabling low-precision operations [17], [18].”
BackgroundMethodSelf-citationSRU-Q: Hardware-friendly Stochastic Rounding Unit-based Gradient Quantization for CNN Training
Jeong, S., Choi, D., & Kim, H. (2024) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas59952.2024.10595985 · Cited by 5
“To solve these problems, various compression studies have been conducted, including pruning [10]–[12], quantization [13], [14], and knowledge distillation [15], [16].”
MethodSelf-citationFeature Distribution-based Knowledge Distillation for Deep Neural Networks
Hong, H., & Kim, H. (2022) · International SoC Design Conference · doi:10.1109/isocc56007.2022.10031412 · Cited by 5
“To address this problem, several CNN compression techniques, including quantization [3], pruning [4], compact model design [5], [6], and knowledge distillation (KD) [7], [8], have been actively studied.”
BackgroundSelf-citationBN-SNN: Spiking neural networks with bistable neurons for object detection
Yasir, S. M., & Kim, H. (2025) · PLoS ONE · doi:10.1371/journal.pone.0327513 · Cited by 4
Self-citationFeature Map-Aware Activation Quantization for Low-bit Neural Networks
Lee, S., & Kim, H. (2021) · International Technical Conference on Circuits/Systems, Computers and Communications (ITC-CSCC) · doi:10.1109/itc-cscc52171.2021.9501414 · Cited by 4
Self-citationKim, H. (2021) · International SoC Design Conference · doi:10.1109/isocc53507.2021.9613926 · Cited by 3
Self-citationPower-Efficient Double-Cyclic Low-Precision Training for Convolutional Neural Networks
Kim, S.-R., & Kim, H. (2022) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas54282.2022.9869948 · Cited by 3
“In CNNs, quantization can be divided into quantization in inference [7],[12],[13],[19]-[23] and training [16]-[18].”
BackgroundMethodSelf-citationPPT-KP: Pruning Point Training-based Kernel Pruning for Deep Convolutional Neural Networks
Koo, K., & Kim, H. (2023) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas57966.2023.10168622 · Cited by 4
Self-citationPerformance Analysis of a Phase-Change Memory System on Various CNN Inference Workloads
Jang, J., Kim, H., & Lee, H. (2022) · International SoC Design Conference · doi:10.1109/isocc56007.2022.10031496 · Cited by 4
“It is noteworthy that the demand for CNN inference in power-constrained edge devices is increasing [6].”
BackgroundSelf-citationBae, S.-H., Lee, H., & Kim, H. (2021) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas51556.2021.9401375 · Cited by 4
Self-citationChoi, D., & Kim, H. (2021) · International SoC Design Conference · doi:10.1109/isocc53507.2021.9613921 · Cited by 4
“In order to solve this problem, research on quantizing the weight (W) and activation (A) parameters of the CNN has been actively conducted [3].”
BackgroundMethodSelf-citationTraffic Light Detection for Autonomous Vehicles in Diverse Environments Through Transfer Learning
Chowdhury, M. R. Z., Seum, A., Talukder, M. R., & Hossain, F. S. (2024) · International Conference on Computer and Information Technology (ICCIT) · doi:10.1109/iccit64611.2024.11022006 · Cited by 2
“To address this, techniques such as pruning, quantization, and knowledge distillation are used to reduce model size and computational requirements [9].”
A Survey and Comparative Analysis of Number Systems for Deep Neural Networks
Alsuhli, G. H., Sakellariou, V., Saleh, H. H., Al‐Qutayri, M., Mohammad, B. S., & Stouraitis, T. (2025) · Proceedings of the IEEE · doi:10.1109/jproc.2025.3578756 · Cited by 7
“Carefully designed quantization techniques can reduce the number of bits to less than 8 [62], [63], [64] and even as low as 2 bits for ternary DNNs (with values {− 1 , 0 , 1 } [65], [66], [67], [68]) or 1 bit for binary DNNs (with values {− 1 , + 1 } [69], [70], [71], [72], [73], [74]).”
Switched 32-Bit Fixed-Point Format for Laplacian-Distributed Data
Denić, B., Perić, Z., Dincic, M. R., et al. (2025) · Information · doi:10.3390/info16070574 · Cited by 2
A Simplified Feature Alignment Strategy for Image Classification Across Domains
Shin, J. W., & Kim, H. (2024) · International Conference on Electronics, Information, and Communication (ICEIC) · doi:10.1109/iceic61013.2024.10457222 · Cited by 2
Self-citationChoi, D. H., & Kim, H. (2024) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas58744.2024.10558001 · Cited by 2
“In particular, quantization approaches that use lower bit widths ( e.g., INT8 and INT4) are gaining attention because of their potential for conserving hardware resources and achieving energy efficiency [10].”
BackgroundSelf-citationChun, D., Lee, H.-J., & Kim, H. (2024) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas59952.2024.10595928 · Cited by 2
“…lightweighting CNN include techniques such as pruning [13]–[17], which involves modification of the architecture to reduce the model size and computational load, and quantization [18]–[20], which adjusts the bit precision of feature maps or parameters to reduce bit operations and memory footprint.”
MethodSelf-citationBranch Convolution Quantization for Object Detection
Li, M., Zhang, F., & Zhang, C. (2024) · Machine Intelligence Research · doi:10.1007/s11633-023-1434-8 · Cited by 3
End-to-End Deep Policy Feedback-Based Reinforcement Learning Method for Quantization in DNNs
Babu, R. L., Gurumoorthy, S., Parameshachari, B. D., Nelson, S. C., & Hua, Q. (2022) · Journal of Circuits Systems and Computers · doi:10.1142/s0218126622502322 · Cited by 2
A Low-Power Reconfigurable DNN Accelerator for Instruction-Extended RISC-V
Wang, H., Li, D., & Isshiki, T. (2024) · IPSJ Transactions on System and LSI Design Methodology · doi:10.2197/ipsjtsldm.17.55 · Cited by 1
Linear Domain-aware Log-scale Post-training Quantization
Kim, S.-R., & Kim, H. (2021) · IEEE International Conference on Consumer Electronics-Asia (ICCE-Asia) · doi:10.1109/icce-asia53811.2021.9642002 · Cited by 2
Self-citationResearch on Navigation and Positioning of Electric Inspection Robot Based on Improved CNN Algorithm
Long, Y., Du, M., Luo, X., Li, S., & Liu, Y. (2022) · Journal of Sensors · doi:10.1155/2022/9369543 · Cited by 1
“It can also directly follow the real-time video and conduct video remote detection synchronously with the machine [11, 12].”
BackgroundDC-MPQ: Distributional Clipping-based Mixed-Precision Quantization for Convolutional Neural Networks
Lee, S., & Kim, H. (2022) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas54282.2022.9869959 · Cited by 1
“This scale factor basically follows the form of the naïve max, min scale factor [16].”
BackgroundSelf-citationObject Classification and Tracking Using Scaled P8 YOLOv4 Lite Model
Shaikh, S., Chopade, J. J., & Kharate, G. K. (2022) · Periodica Polytechnica Electrical Engineering and Computer Science · doi:10.3311/ppee.20685 · Cited by 1
“In order to address the issues with floating-point-based quantization methods for YOLOv3 and YOLOv4, Kim and Kim [8] sug - gested a fixed-point-based quantization technique tailored for embedded platforms.”
MethodConventional Number Systems for DNN Architectures
Alsuhli, G. H., Sakellariou, V., Saleh, H. H., Al‐Qutayri, M., Mohammad, B. S., & Stouraitis, T. (2023) · Synthesis lectures on engineering, science, and technology · doi:10.1007/978-3-031-38133-1_3 · Cited by 1
Adaptive Quantization with Mixed-Precision Based on Low-Cost Proxy
Chen, J., Yang, Q., Tian, S., & Zhang, S. (2024) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp48485.2024.10447866 · Cited by 4
RETRACTED: Detection of an in-housed pig using modified YOLOv5 model
Devi, S. J., Doley, J., & Gupta, V. K. (2024) · Journal of Intelligent & Fuzzy Systems · doi:10.3233/jifs-231032 · Cited by 1
RetractedEnhancing Pseudo-Labeling Performance in Object Detection Using Gaussian Mixture Modeled Uncertainty
Lee, S., & Kim, H. (2024) · International Conference on Electronics, Information, and Communication (ICEIC) · doi:10.1109/iceic61013.2024.10457103 · Cited by 2
“However, achieving high performance requires not only an enormous number of parameters and computations [4], [5], but also large-scale training data, where accurate labeling demands substantial cost and time.”
BackgroundSelf-citationIdentification of insect pests and prediction of pesticides using YOLOv4 deploying onto UbiOps
Devi, K. J., Chanu, M. M., Challa, G., et al. (2025) · AIP conference proceedings · doi:10.1063/5.0247101 · Cited by 1
Hong, H., & Kim, H. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.111597 · Cited by 3
Self-citationObject Detector Friendly Frame Memory Compression Based on 1-D DWT and SPIHT
Shin, J. W., & Kim, H. (2021) · International Technical Conference on Circuits/Systems, Computers and Communications (ITC-CSCC) · doi:10.1109/itc-cscc52171.2021.9501261
Self-citationAnalysis for Implementing Power-Efficient Convolutional Operators on FPGA Platforms
Park, J., & Kim, H. (2021) · IEEE International Conference on Consumer Electronics-Asia (ICCE-Asia) · doi:10.1109/icce-asia53811.2021.9641973
Self-citationAnalysis of Hardware Prefetchers Suitable for CNN Applications
Seong, H. G., Lee, H., Kim, H., & Lee, H. (2021) · IEEE International Conference on Consumer Electronics-Asia (ICCE-Asia) · doi:10.1109/icce-asia53811.2021.9641920
Self-citationActivation Distribution-based Layer-wise Quantization for Convolutional Neural Networks
Ki, S., & Kim, H. (2022) · International Conference on Electronics, Information, and Communication (ICEIC) · doi:10.1109/iceic54506.2022.9748745
Self-citationGPIL: Gradient with PseudoInverse Learning for High Accuracy Fine-Tuning
Lee, G., Kim, N. J., & Kim, H. (2023) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas57966.2023.10168584 · Cited by 2
“In particular, many network compression studies, including pruning [11], [12] and quantization [13], [14], have been conducted to practically utilize CNNs in resource-constrained environments, such as mobile platforms.”
BackgroundSelf-citationWu, Q. (2023) · International Journal of Computational Intelligence and Applications · doi:10.1142/s1469026823500232
Decoupling efficiency-performance optimization for modern neural networks
Jin, Q. (2023) · Digital Repository Service (Northeastern University) · doi:10.17760/d20592335
Koo, K., & Kim, H. (2024) · International Conference on Electronics, Information, and Communication (ICEIC) · doi:10.1109/iceic61013.2024.10457126
“Therefore, to practically deploy advanced networks, compression becomes essential [7], [8].”
BackgroundSelf-citationChoi, D. H., & Kim, H. (2024) · International Conference on Electronics, Information, and Communication (ICEIC) · doi:10.1109/iceic61013.2024.10457260
Self-citationHardware-aware Network Compression for Hybrid Vision Transformer via Low-Rank Approximation
Kang, B. J., Kim, N. J., & Kim, H. (2024) · International SoC Design Conference · doi:10.1109/isocc62682.2024.10762117
“To solve this problem, various model compression methods have been proposed to reduce the size of the model parameters (e.g., quantization, pruning, and low-rank approximation) [8]-[10].”
Self-citationOptimizing Deep Neural Network Precision for Processing-in-Memory: A Memory Bottleneck Perspective
Hwang, I., Jang, J., & Kim, H. (2025) · International Conference on Electronics, Information, and Communication (ICEIC) · doi:10.1109/iceic64972.2025.10879758
Self-citationGASQ: Hardware-Friendly Gradient Distribution-Aware Split Quantization for Low-Bit CNN Training
Jeong, S., Koo, K., & Kim, H. (2025) · International Conference on Electronics, Information, and Communication (ICEIC) · doi:10.1109/iceic64972.2025.10879630
Self-citationSynthèse de code virgule fixe pour les réseaux de neurones
Benmaghnia, H. (2022) · theses.fr (ABES) · doi:10.70675/27a5d0d4z766dz4b6fz9953z0e62b444668a
Importance-Aware Quantization Strategy for Layer-Sensitive YOLO Architectures
Valadão, M. D. M., Pessoa, L. D. S., Pantoja, M. S., Cavalcante, V. R. M., Corrêa, A. J. P., Melo, D. S., de Mattos, E. U., Bessa, A. R. R., Cabral, I. A., Amoedo, D., et al. (2026) · IEEE International Conference on Consumer Technology (ICCT-Pacific) · doi:10.1109/icct-pacific69083.2026.11518766
“The third stage introduces an optimization loop that refines the quantization levels { l i } across layers to jointly satisfy a target accuracy and an average bitwidth objective [11].”
Jeong, S., Lee, S., & Kim, H. (2026) · Neurocomputing · doi:10.1016/j.neucom.2026.134893
Self-citation
Metadata
- Authors
- Kim, S., & Kim, H.
- Affiliations
- Seoul National University of Science and Technology (KR)
- Venue
- IEEE Access
- Published
- 2021
- Volume
- 9
- Pages
- 20828-20839
- ISSN
- 2169-3536 (ISSN-L)
- OpenAlex
- W3122390253
- Cited by
- 71
- References
- 63
- Citations by year
- 2026: 3 · 2025: 8 · 2024: 22 · 2023: 11 · 2022: 15 · 2021: 11
- FWCI
- 4.52
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Locations
- IEEE Access (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- Institute of Information & Communications Technology Planning & Evaluation (IITP) Grant funded by the Korean Government (MSIT) (2020-0-01305)
- Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (NRF-2019R1A6A1A03032119)
- 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 › Computational Theory and Mathematics › Numerical Methods and Algorithms
- Keywords
- fixed-point quantization, object detection, model compression, YOLOv3, YOLOv4, weight distribution, embedded hardware acceleration, low-bit precision, accuracy degradation
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (qwen3.8-omni-flash and 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
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.