Research map / Detection and segmentation / CNN features
Task-Aware Image Signal Processor for Advanced Visual Perception
Kai Chen, Jin Xiao, Leheng Zhang, Kexuan Shi, Shuhang Gu
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 Global Luminance Calibration (ablation) | TA-ISP baseline · LOD | AP: 58.5 -> 60.4 | Improved |
| add Hierarchical Spatial Attention (ablation) | TA-ISP with GLC · LOD | AP: 60.4 -> 63.0 | Improved |
| add Region-Guided Feature Conditioning (ablation) | TA-ISP with GLC+HSA · LOD | AP: 63.0 -> 63.9 | Improved |
| add Global Luminance Calibration (ablation) | TA-ISP baseline · ROD | AP: 36.1 -> 36.3 | No clear effect |
| add Hierarchical Spatial Attention (ablation) | TA-ISP with GLC · ROD | AP: 36.3 -> 36.8 | Improved |
| add Region-Guided Feature Conditioning (ablation) | TA-ISP with GLC+HSA · ROD | AP: 36.8 -> 38.0 | Improved |
| increase mask layers from 8 to 16 (ablation) | TA-ISP · LOD | AP: 63.3 -> 63.9 | Improved |
| increase mask layers from 16 to 32 (ablation) | TA-ISP · LOD | AP: 63.9 -> 62.0 | Hurt |
| increase mask layers from 16 to 32 (ablation) | TA-ISP · PASCAL RAW | AP: 89.9 -> 88.4 | Hurt |
| use TA-ISP instead of RAW-Adapter | RAW-Adapter · LOD | AP: 62.1 -> 63.9 | Improved |
| use TA-ISP instead of demosaicing under 10% training data | Demosaic · PASCAL RAW limited data | AP: 73.9 -> 74.1 | Improved |
| train existing ISP methods on 10% data (MW-ISPNet vs demosaic) | MW-ISPNet · PASCAL RAW limited data | AP: 73.9 -> 68.4 | Hurt |
Closest papers by meaning
- AdaptiveISP: Learning an Adaptive Image Signal Processor for Object Detection (NeurIPS 2024 · similarity 0.88)
- Dark-ISP: Enhancing RAW Image Processing for Low-Light Object Detection (ICCV 2025 · similarity 0.87)
- Beyond RGB: Adaptive Parallel Processing for RAW Object Detection (ICCV 2025 · similarity 0.87)
- RL-SeqISP: Reinforcement Learning-Based Sequential Optimization for Image Signal Processing (AAAI 2024 · similarity 0.84)
- RGB-Event ISP: The Dataset and Benchmark (ICLR 2025 · similarity 0.83)
- Transtreaming: Adaptive Delay-aware Transformer for Real-time Streaming Perception (AAAI 2025 · 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.
