Progressive Learning of Low-Precision Networks for Image Classification

Zhou, Ziheng; Zhou, Wengang; Lv, Xutao; Huang, Xuan; Wang, Xiaoyu; Li, Houqiang

doi:10.1109/tmm.2020.2990087

Cited by 7 publications

(3 citation statements)

References 39 publications

(88 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…is is also the direct principle of teaching resource classification. When establishing a classification system, artificial intelligence teaching resources should consider the needs of different users, make artificial intelligence teaching resources serve well, and follow the user-oriented principle [9]. Artificial intelligence teaching resources should have a classification system that meets the needs of information development, with enough space reserved in the classification system for new knowledge to be added to meet people's needs for new knowledge and new disciplines.…”

Section: Construction Of Aerobics Course Online Teachingmentioning

confidence: 99%

Classification Method of Aerobics Course Online Teaching Resources Based on Artificial Intelligence Technology

Zhao

Zhang

Wang

et al. 2022

Wireless Communications and Mobile Computing

View full text Add to dashboard Cite

In light of the current uneven distribution of aerobics course online resources, this paper presents research on the classification method of aerobics course online teaching resources based on artificial intelligence technology, constructs the display education resource management system based on artificial intelligence technology, realizes the classification of teaching information characteristics, and constructs the classification and evaluation algorithm of aerobiology course online resources. The online teaching resource classification of aerobics course achieves its design goal. Finally, the experiment shows that the artificial intelligence-based online teaching resource classification method for aerobics course is highly practicable and fully meets the research requirements.

show abstract

Section: Construction Of Aerobics Course Online Teachingmentioning

confidence: 99%

Classification Method of Aerobics Course Online Teaching Resources Based on Artificial Intelligence Technology

Zhao

Zhang

Wang

et al. 2022

Wireless Communications and Mobile Computing

View full text Add to dashboard Cite

show abstract

“…Different with [44] whose clipping value is fixed, Choi et al [47] optimized the clipping range of activations through training and then linearly quantized both weights and activations to 4-bit. Zhou et al [48] equipped a low-precision network with a full-precision network and then gradually remove the impact of the full-precision network during the training.…”

Section: B Neural Network Quantizationmentioning

confidence: 99%

End-To-End Learned Image Compression With Fixed Point Weight Quantization

Sun

Cheng

Takeuchi

et al. 2020

2020 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

End-to-end Learned image compression (LIC) has reached the traditional hand-crafted methods such as BPG (HEVC intra) in terms of the coding gain. However, the large network size prohibits the usage of LIC on resource-limited embedded systems. This paper reduces the network complexity by quantizing both weights and activations. 1) For the weight quantization, we study different kinds of grouping and quantization scheme at first. A channel-wise non-linear quantization scheme is determined based on the coding gain analysis. After that, we propose a fine tuning scheme to clip the weights within a certain range so that the quantization error can be reduced. 2) For the activation quantization, we first propose multiple nonlinear quantization codebooks with different maximum dynamic ranges. By selecting an optimal one through a multiplexer, the quantization range can be saturated to the greatest extent. In addition, we also exploit the mean-removed quantization for the analysis transform outputs in order to reduce the bit-width cost for the specific channel with the large non-zero mean. By quantizing each weight and activation element from 32-bit floating point to 8-bit fixed point, the memory cost for both weight and activation can be reduced by 75% with negligible coding performance loss. As a result, our quantized LIC can still outperform BPG in terms of MS-SSIM. To our best knowledge, this is the first work to give a complete analysis on the coding gain and the memory cost for a quantized LIC network, which validates the feasibility of the hardware implementation.

show abstract

“…Different with [31] whose clipping value is fixed, Choi et al [34] optimized the clipping range of activations through training and then linearly quantized both weights and activations to 4-bit. Zhou et al [35] equipped a low-precision network with a full-precision network and then gradually remove the impact of the full-precision network during the training. [36] proposed an element-wise gradient scaling to replace the straight-through-estimator in the training.…”

Section: Introductionmentioning

confidence: 99%

Q-LIC: Quantizing Learned Image Compression with Channel Splitting

Sun¹,

Lei²,

Katto³

2022

Preprint

View full text Add to dashboard Cite

Learned image compression (LIC) has reached a comparable coding gain with traditional hand-crafted methods such as VVC intra. However, the large network complexity prohibits the usage of LIC on resource-limited embedded systems. Network quantization is an efficient way to reduce the network burden. This paper presents a quantized LIC (QLIC) by channel splitting. First, we explore that the influence of quantization error to the reconstruction error is different for various channels. Second, we split the channels whose quantization has larger influence to the reconstruction error. After the splitting, the dynamic range of channels is reduced so that the quantization error can be reduced. Finally, we prune several channels to keep the number of overall channels as origin. By using the proposal, in the case of 8-bit quantization for weight and activation of both main and hyper path, we can reduce the BD-rate by 0.61%-4.74% compared with the previous QLIC. Besides, we can reach better coding gain compared with the state-of-the-art network quantization method when quantizing MS-SSIM models. Moreover, our proposal can be combined with other network quantization methods to further improve the coding gain. The moderate coding loss caused by the quantization validates the feasibility of the hardware implementation for QLIC in the future.

show abstract

Progressive Learning of Low-Precision Networks for Image Classification

Cited by 7 publications

References 39 publications

Classification Method of Aerobics Course Online Teaching Resources Based on Artificial Intelligence Technology

Classification Method of Aerobics Course Online Teaching Resources Based on Artificial Intelligence Technology

End-To-End Learned Image Compression With Fixed Point Weight Quantization

Q-LIC: Quantizing Learned Image Compression with Channel Splitting

Contact Info

Product

Resources

About