Research map / Architectures and efficiency / CNN design
Wavelet Feature Maps Compression for Image-to-Image CNNs
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| WCC 50% shrinkage on 8-bit activations vs 4-bit activation quantization at similar BOPs | EfficientDet-D1 · MS COCO val | mAP: 24.70 -> 31.15 | Helped |
| add 50% wavelet shrinkage to 4bit/8bit baseline (ablation) | EfficientDet-D1 · MS COCO val | mAP: 31.44 -> 31.15 | Hurt |
| reduce shrinkage from 50% to 25% (ablation) | WCC EfficientDet-D1 · MS COCO val | mAP: 31.15 -> 27.49 | Hurt |
| add 50% WCC to baseline (ablation) | DeepLabV3plus · Pascal VOC val | mIoU: 0.675 -> 0.661 | Hurt |
| reduce shrinkage to 12.5% (ablation) | WCC DeepLabV3plus · Pascal VOC val | mIoU: 0.661 -> 0.515 | Hurt |
| use WCC 25% instead of 8bit/2bit activation quantization | Monodepth2 · KITTI | AbsRel: 0.268 -> 0.112 | Helped |
| WCC 50% vs 8bit/4bit quantization | Monodepth2 · KITTI | AbsRel: 0.097 -> 0.098 | No clear effect |
| add 50% WCC to 8bit/8bit EDSR (ablation) | EDSR · DIV2K 2x SR | PSNR: 34.53 -> 34.53 | No clear effect |
| shrink FP32 EDSR with WCC 12.5% (ablation) | EDSR · DIV2K 2x SR | PSNR: 35.02 -> 34.76 | Hurt |
Closest papers by meaning
- Network Memory Footprint Compression Through Jointly Learnable Codebooks and Mappings (ICLR 2024 · similarity 0.85)
- Quantization without Tears (CVPR 2025 · similarity 0.84)
- Another Way to the Top: Exploit Contextual Clustering in Learned Image Coding (AAAI 2024 · similarity 0.84)
- Frequency-Biased Synergistic Design for Image Compression and Compensation (CVPR 2025 · similarity 0.84)
- When MLLMs Meets Compression Distortion: A Coding Paradigm Tailored to MLLMs (ICLR 2026 · similarity 0.84)
- Benchmarking and Enhancing VLM for Compressed Image Understanding (ICML 2026 · similarity 0.83)
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.