Research map / Image generation / Autoregressive and latent generation
Focal Plane Visual Feature Generation and Matching on a Pixel Processor Array
Hongyi Zhang, Laurie Bose, Jianing Chen, Piotr Dudek, Walterio Mayol-Cuevas
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 |
|---|---|---|---|
| use 128-bit instead of 64-bit descriptors (ablation) | PPA digital matching · Oxford Robotic Car Img2Img | accuracy (%): 90.9 -> 97.8 | Improved |
| use 256-bit instead of 128-bit descriptors (ablation) | PPA digital matching · Oxford Robotic Car Img2Img | accuracy (%): 97.8 -> 98.5 | Improved |
| add cross check test (ablation) | PPA analogue 64-bit · Oxford Robotic Car Img2Img | accuracy (%): 78.6 -> 83.6 | Improved |
| add cross check test (ablation) | PPA digital 128-bit · Oxford Robotic Car Img2Img | accuracy (%): 97.8 -> 98.1 | Improved |
| add cross check test (ablation) | PPA digital 64-bit · Oxford Robotic Car Seq2Seq | accuracy (%): 98.3 -> 98.2 | No clear effect |
| use digital instead of analogue bit summation (ablation) | PPA 64-bit · Oxford Robotic Car Img2Img | accuracy (%): 78.6 -> 90.9 | Improved |
| use digital instead of analogue bit summation (ablation) | PPA 128-bit · Oxford Affine Wall pair 1|2 | recognition rate (%): 78.5 ± 2.7 -> 94.0 ± 1.0 | Improved |
| use digital instead of analogue bit summation (ablation) | PPA 128-bit · Oxford Affine Wall pair 1|4 | recognition rate (%): 45.3 ± 3.3 -> 81.7 ± 2.9 | Improved |
| run on PPA instead of PC (analogue noise) (ablation) | BRIEF 128-bit digital · Oxford Affine Leuven pair 1|2 | recognition rate (%): 100.0 ± 0.0 -> 95.4 ± 1.0 | Hurt |
| apply optimized pixel shifting vs naive | BRIEF descriptor generation 256-bit · analogue data transfer operations | operations: 4852 -> 1902 | Improved |
| apply optimized pixel shifting vs naive | BRIEF descriptor generation 64-bit · analogue data transfer operations | operations: 1224 -> 632 | Improved |
| increase descriptor size from 64 to 256 bits (ablation) | PPA digital matching · frame rate | fps: 209 -> 50 | Hurt |
Closest papers by meaning
- Descriptor-In-Pixel : Point-Feature Tracking For Pixel Processor Arrays (CVPR 2025 · similarity 0.92)
- Visual Odometry for Pixel Processor Arrays (ICCV 2017 · similarity 0.87)
- Fully Embedding Fast Convolutional Networks on Pixel Processor Arrays (ECCV 2020 · similarity 0.87)
- A Camera That CNNs: Towards Embedded Neural Networks on Pixel Processor Arrays (ICCV 2019 · similarity 0.86)
- PixVOD: Pixel-Distributed Direct Visual Odometry and Depth Estimation (ECCV 2026 · similarity 0.82)
- Towards Complementary Keypoint Detection via Mixture of Detectors (NeurIPS 2026 · similarity 0.81)
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.
