Research map / Detection and segmentation / CNN features
The Differentiable Lens: Compound Lens Search Over Glass Surfaces and Materials for Object Detection
Geoffroi Côté, Fahim Mannan, Simon Thibault, Jean-François Lalonde, Felix Heide
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 |
|---|---|---|---|
| Jointly optimize the lens and detector instead of fine-tuning the detector with a fixed lens | Doublet · BDD100K, 1× and 2× resolution; Udacity, 1× and 2× resolution | AP: 30.3 -> 32.0; 25.0 -> 28.1; 23.5 -> 25.6; 18.4 -> 21.9 | Improved |
| Jointly optimize the lens and detector instead of fine-tuning the detector with a fixed lens | Cooke · BDD100K and Udacity, 1× resolution | AP: 33.0 -> 33.3; 26.3 -> 26.8 | Improved |
| Jointly optimize the lens and detector instead of fine-tuning the detector with a fixed lens | Cooke · BDD100K and Udacity, 2× resolution | AP: 31.5 -> 31.7; 24.9 -> 24.8 | Mixed |
| Jointly optimize the lens and detector instead of fine-tuning the detector with a fixed lens | Tessar · BDD100K and Udacity, 1× and 2× resolution | AP: 33.4 -> 33.6; 31.3 -> 32.2; 26.9 -> 27.0; 24.4 -> 26.2 | Improved |
| Use quantized continuous rather than continuous glass variables (ablation) | Jointly optimized Tessar · BDD100K, 2× resolution | AP: 25.9 -> 32.2 | Improved |
| Add a paraxial image solve (ablation) | Jointly optimized Tessar · BDD100K, 2× resolution | AP: 26.5 -> 32.2 | Improved |
| Add ray path loss (ablation) | Jointly optimized Tessar · BDD100K, 2× resolution | AP: 15.4 -> 32.2 | Improved |
| Add ray angle loss (ablation) | Jointly optimized Tessar · BDD100K, 2× resolution | AP: 24.2 -> 32.2 | Improved |
| Add spot size loss (ablation) | Jointly optimized Tessar · BDD100K, 2× resolution | AP: 11.1 -> 32.2 | Improved |
Closest papers by meaning
- Lens Component Deletion based on Differentiable Ray Tracing (CVPR 2026 · similarity 0.85)
- A Differentiable Wave Optics Model for End-to-End Computational Imaging System Optimization (ICCV 2025 · similarity 0.85)
- Rich Context Aggregation With Reflection Prior for Glass Surface Detection (CVPR 2021 · similarity 0.85)
- Multi-View Dynamic Reflection Prior for Video Glass Surface Detection (AAAI 2024 · similarity 0.84)
- LensDesigner: A Self-Improving Agent for Optical Lens Design (NeurIPS 2026 · similarity 0.83)
- A Physics-informed Low-rank Deep Neural Network for Blind and Universal Lens Aberration Correction (CVPR 2024 · similarity 0.82)
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.
