Research map / Label-efficient and robust learning / Self-supervised and contrastive
Image Coding for Machines with Omnipotent Feature Learning
Ruoyu Feng, Xin Jin, Zongyu Guo, Runsen Feng, Yixin Gao, Tianyu He, Zhizheng Zhang, Simeng Sun, Zhibo Chen
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 |
|---|---|---|---|
| add Information Filtering module with entropy constraint (ablation) | Omni-ICM · PASCAL VOC object detection | rate-accuracy | Improved |
| add feature-level distortion loss to feature compression (ablation) | Omni-ICM · PASCAL VOC object detection | rate-accuracy | Improved |
| compress omnipotent features instead of task-specific features or reconstructed images | Omni-ICM vs VVC · Cityscapes semantic segmentation | BD-rate: -24.6 -> -72.0 | Improved |
| compress omnipotent features instead of task-specific features or reconstructed images | Omni-ICM vs cheng+Joint · COCO object detection | BD-rate: -3.0 -> -43.9 | Improved |
| compress omnipotent features instead of using RoI bit allocation | Omni-ICM vs VVC+RoI · PASCAL VOC object detection | BD-rate: -32.9 -> -35.1 | Improved |
| use contrastive pretraining instead of supervised ImageNet pretraining for task models | Task models · multiple tasks | task performance | Improved |
| compress omnipotent features instead of uncompressed contrastive backbone | Omni-ICM · uncompressed baseline | task performance | Hurt |
| use Omni-ICM at highest bitrate | Omni-ICM · COCO pose estimation | AP: more than 2 points gap | Hurt |
Closest papers by meaning
- Unified Coding for Both Human Perception and Generalized Machine Analytics with CLIP Supervision (AAAI 2025 · similarity 0.89)
- Video Coding for Machines: Compact Visual Representation Compression for Intelligent Collaborative Analytics (TPAMI 2024 · similarity 0.87)
- Diff-ICMH: Harmonizing Machine and Human Vision in Image Compression with Generative Prior (NeurIPS 2025 · similarity 0.86)
- Test-Time Fine-Tuning of Image Compression Models for Multi-Task Adaptability (CVPR 2025 · similarity 0.84)
- All-in-One Image Coding for Joint Human-Machine Vision with Multi-Path Aggregation (NeurIPS 2024 · similarity 0.84)
- When MLLMs Meets Compression Distortion: A Coding Paradigm Tailored to MLLMs (ICLR 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.
