Subset-Selection Weight Post-Training Quantization Method for Learned Image Compression Task
Post-training quantization(PTQ) has been widely studied in recent years because it does not require retraining the network or the entire training dataset. However, naively applying the PTQ method to quantize low-level image tasks such as learned image compression(LIC) usually incurs significant accu...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10819398/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841542518054846464 |
|---|---|
| author | Jinru Yang Xiaoqin Wang Qiang Li Shushan Qiao Yumei Zhou |
| author_facet | Jinru Yang Xiaoqin Wang Qiang Li Shushan Qiao Yumei Zhou |
| author_sort | Jinru Yang |
| collection | DOAJ |
| description | Post-training quantization(PTQ) has been widely studied in recent years because it does not require retraining the network or the entire training dataset. However, naively applying the PTQ method to quantize low-level image tasks such as learned image compression(LIC) usually incurs significant accuracy degradation. Existing solutions often aim to optimize the calibration process or minimize quantization loss in the context of uniform quantization, which makes it difficult to further reduce the quantization loss. This work achieves accurate weight quantization using a non-uniform quantization method called subset-selection within the PTQ method. Subset-selection method uses a clustering algorithm to select a proper batch of quantization points from a large uniformly constructed quantization points pool that matches the current layer’s weight distribution correctly. Compared to the state-of-the-art LIC PTQ method, experimental results show that our proposed method achieves an average 0.8% better BD-rate on several LIC models. Meanwhile, we extended our method to image classification models and achieved an average 0.64% better accuracy, further proving the generalization of our method. |
| format | Article |
| id | doaj-art-048a9fdf571142e1ab268396854c2fca |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-048a9fdf571142e1ab268396854c2fca2025-01-14T00:02:37ZengIEEEIEEE Access2169-35362025-01-01135145515310.1109/ACCESS.2024.352455310819398Subset-Selection Weight Post-Training Quantization Method for Learned Image Compression TaskJinru Yang0https://orcid.org/0009-0002-7500-9715Xiaoqin Wang1https://orcid.org/0000-0001-5890-8079Qiang Li2Shushan Qiao3https://orcid.org/0000-0002-9102-2111Yumei Zhou4Institute of Microelectronics of the Chinese Academy of Sciences, Chaoyang, Beijing, ChinaInstitute of Microelectronics of the Chinese Academy of Sciences, Chaoyang, Beijing, ChinaInstitute of Microelectronics of the Chinese Academy of Sciences, Chaoyang, Beijing, ChinaInstitute of Microelectronics of the Chinese Academy of Sciences, Chaoyang, Beijing, ChinaInstitute of Microelectronics of the Chinese Academy of Sciences, Chaoyang, Beijing, ChinaPost-training quantization(PTQ) has been widely studied in recent years because it does not require retraining the network or the entire training dataset. However, naively applying the PTQ method to quantize low-level image tasks such as learned image compression(LIC) usually incurs significant accuracy degradation. Existing solutions often aim to optimize the calibration process or minimize quantization loss in the context of uniform quantization, which makes it difficult to further reduce the quantization loss. This work achieves accurate weight quantization using a non-uniform quantization method called subset-selection within the PTQ method. Subset-selection method uses a clustering algorithm to select a proper batch of quantization points from a large uniformly constructed quantization points pool that matches the current layer’s weight distribution correctly. Compared to the state-of-the-art LIC PTQ method, experimental results show that our proposed method achieves an average 0.8% better BD-rate on several LIC models. Meanwhile, we extended our method to image classification models and achieved an average 0.64% better accuracy, further proving the generalization of our method.https://ieeexplore.ieee.org/document/10819398/Convolutional neural networkpost-training quantizationnon-uniform quantizationlearned image compression |
| spellingShingle | Jinru Yang Xiaoqin Wang Qiang Li Shushan Qiao Yumei Zhou Subset-Selection Weight Post-Training Quantization Method for Learned Image Compression Task IEEE Access Convolutional neural network post-training quantization non-uniform quantization learned image compression |
| title | Subset-Selection Weight Post-Training Quantization Method for Learned Image Compression Task |
| title_full | Subset-Selection Weight Post-Training Quantization Method for Learned Image Compression Task |
| title_fullStr | Subset-Selection Weight Post-Training Quantization Method for Learned Image Compression Task |
| title_full_unstemmed | Subset-Selection Weight Post-Training Quantization Method for Learned Image Compression Task |
| title_short | Subset-Selection Weight Post-Training Quantization Method for Learned Image Compression Task |
| title_sort | subset selection weight post training quantization method for learned image compression task |
| topic | Convolutional neural network post-training quantization non-uniform quantization learned image compression |
| url | https://ieeexplore.ieee.org/document/10819398/ |
| work_keys_str_mv | AT jinruyang subsetselectionweightposttrainingquantizationmethodforlearnedimagecompressiontask AT xiaoqinwang subsetselectionweightposttrainingquantizationmethodforlearnedimagecompressiontask AT qiangli subsetselectionweightposttrainingquantizationmethodforlearnedimagecompressiontask AT shushanqiao subsetselectionweightposttrainingquantizationmethodforlearnedimagecompressiontask AT yumeizhou subsetselectionweightposttrainingquantizationmethodforlearnedimagecompressiontask |