Research map / Architectures and efficiency / Quantization
TexQ: Zero-shot Network Quantization with Texture Feature Distribution Calibration
Xinrui Chen, Yizhi Wang, Renao Yan, Yiqing Liu, Tian Guan, Yonghong He
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| remove both L_LAWS and L_L-BNS (only MKD kept) (ablation) | TexQ · ImageNet W3A3 ResNet-18 | top-1 accuracy (%): 50.94 -> 42.57 | Hurt |
| add calibration losses L_LAWS and L_L-BNS together (ablation) | TexQ · ImageNet W3A3 ResNet-18 | top-1 accuracy (%): 8.37% loss when removed | Improved |
| drop one or two of the three modules (ablation) | TexQ · ImageNet W3A3 ResNet-18 | top-1 accuracy (%): 50.94 -> 49.03 | Hurt |
| replace prior best AdaDFQ with TexQ at 3-bit | ResNet-18 · ImageNet W3A3 | top-1 accuracy (%): 38.10 -> 50.28 | Improved |
| replace AdaDFQ with TexQ at 4-bit | ResNet-18 · ImageNet W4A4 | top-1 accuracy (%): 66.53 -> 67.73 | Improved |
| replace AdaDFQ with TexQ at 3-bit | ResNet-50 · ImageNet W3A3 | top-1 accuracy (%): 17.63 -> 25.27 | Improved |
| replace AdaDFQ with TexQ at 3-bit | ResNet-20 · CIFAR-10 W3A3 | top-1 accuracy (%): 84.89 -> 86.47 | Improved |
| replace AdaDFQ with TexQ at 3-bit | ResNet-20 · CIFAR-100 W3A3 | top-1 accuracy (%): 52.74 -> 55.87 | Improved |
| replace AdaDFQ with TexQ at 4-bit | ResNet-20 · CIFAR-10 W4A4 | top-1 accuracy (%): 92.31 -> 92.68 | Improved |
| TexQ vs real-data fine-tuning with same framework | ResNet-20 · CIFAR-10 W4A4 | top-1 accuracy (%): 91.52 -> 92.68 | Improved |
| grid-search texture hyperparameters k, theta_U, theta_L, epsilon (ablation) | TexQ · ImageNet W3A3 ResNet-18 | top-1 accuracy (%): best 50.94 | Mixed |
| mixup probability p=20% for MKD (ablation) | TexQ · ImageNet W3A3 | top-1 accuracy (%) | Improved |
Later papers that compared with this method
3 later accepted papers report TexQ as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Accurate Zero-shot Quantization via Hierarchical Teacher-Assistant Distillation (ECCV 2026 · better on 3 of 3 shared results)
- Enhancing Diversity for Data-free Quantization (CVPR 2025 · better on 6 of 6 shared results)
- GenQ: Quantization in Low Data Regimes with Generative Synthetic Data (ECCV 2024 · better on 4 of 4 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Qimera — Qimera: Data-free Quantization with Synthetic Boundary Supporting Samples (NeurIPS 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ZeroQ: A Novel Zero Shot Quantization Framework (CVPR 2020 · similarity 0.89)
- SynQ: Accurate Zero-shot Quantization by Synthesis-aware Fine-tuning (ICLR 2025 · similarity 0.88)
- IntraQ: Learning Synthetic Images With Intra-Class Heterogeneity for Zero-Shot Network Quantization (CVPR 2022 · similarity 0.86)
- Sharpness-Aware Data Generation for Zero-shot Quantization (ICML 2024 · similarity 0.85)
- 2DQuant: Low-bit Post-Training Quantization for Image Super-Resolution (NeurIPS 2024 · similarity 0.84)
- PTMQ: Post-training Multi-Bit Quantization of Neural Networks (AAAI 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
