Research map / Detection and segmentation / CNN features
Dark-ISP: Enhancing RAW Image Processing for Low-Light Object Detection
Jiasheng Guo, Xin Gao, Yuxiang Yan, Guanghao Li, Jian Pu
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 |
|---|---|---|---|
| feed Bayer RAW through Dark-ISP instead of default ISP | RetinaNet ResNet50 · LOD (Bayer RAW) | mAP: 67.3 -> 70.4 | Improved |
| use Bayer RAW input instead of RGB for default ISP | RetinaNet ResNet50 · LOD | mAP: 59.1 -> 67.3 | Improved |
| add a pre-trained denoising stage (SID, two-stage) | RetinaNet ResNet50 · LOD (Bayer RAW) | mAP: 67.3 -> 64.7 | Hurt |
| apply Dark-ISP on NOD Nikon | RetinaNet ResNet50 · NOD Nikon | mAP75: 30.8 -> 30.7 | No clear effect |
| add Self-Boost Regularization to linear plus nonlinear components (ablation) | Dark-ISP · LOD (Bayer RAW) | mAP: 68.7 -> 70.4 | Improved |
| combine linear and nonlinear components instead of nonlinear only (ablation) | Dark-ISP · LOD (Bayer RAW) | mAP: 67.1 -> 68.7 | Improved |
| use nonlinear component alone versus linear alone (ablation) | Dark-ISP · LOD (Bayer RAW) | mAP: 66.6 -> 67.1 | Improved |
| use polynomial-basis nonlinear component instead of Zero-DCE (ablation) | Dark-ISP · LOD (Bayer RAW) | mAP: 68.0 -> 70.4 | Improved |
| add skip connection in the nonlinear component (ablation) | Dark-ISP · LOD (Bayer RAW) | mAP: 68.6 -> 70.4 | Improved |
| use polynomial basis instead of fixed gamma correction (ablation) | Dark-ISP · LOD (Bayer RAW) | mAP: 66.4 -> 70.4 | Improved |
| apply Dark-ISP on synthetic SynCOCO | RetinaNet ResNet50 · SynCOCO (Bayer RAW) | mAP: 21.5 -> 23.1 | Improved |
Closest papers by meaning
- ISP-Teacher:Image Signal Process with Disentanglement Regularization for Unsupervised Domain Adaptive Dark Object Detection (AAAI 2024 · similarity 0.90)
- AdaptiveISP: Learning an Adaptive Image Signal Processor for Object Detection (NeurIPS 2024 · similarity 0.88)
- Boosting Object Detection with Zero-Shot Day-Night Domain Adaptation (CVPR 2024 · similarity 0.86)
- RAWild: Toward Sensor-Agnostic RAW Object Detection via Physics-Guided Curve and Grid Modeling (NeurIPS 2026 · similarity 0.86)
- FRBNet: Revisiting Low-Light Vision through Frequency-Domain Radial Basis Network (NeurIPS 2025 · similarity 0.85)
- Guiding a Harsh-Environments Robust Detector via RAW Data Characteristic Mining (AAAI 2024 · similarity 0.85)
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.
