Research map / Probabilistic and scientific modeling / Molecules and proteins
How Do Neural Networks Estimate Optical Flow A Neuropsychology-Inspired Study
David Benjamin De Jong, Federico Paredes-Vallés, Guido de Croon
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 |
|---|---|---|---|
| shrink receptive field of c6 by using 1x1 kernels (FlowNetXS 255 px, FlowNetXXS 191 px vs 383 px) (ablation) | FlowNetS · diagonally translating bars | EPE vs bar scale: 383 -> 255 -> 191 pixels receptive field | Hurt |
| go deeper in the encoder (c1-c2 vs c3-c6) | FlowNetS Gabor fit · Gabor spectral fitting | L_norm: 4.71 (conv1) -> 1.42 (conv5_1) | Mixed |
| go from conv5_1 to conv6_1 | FlowNetS Gabor fit · Gabor spectral fitting | L_norm: 1.42 -> 1.65 | Hurt |
| use ReLU, 1x1 bias-free pf layers and 300K iterations instead of original FlowNetS | FlowNetS · FlyingChairs test | EPE: 2,71 -> 3,10 | Hurt |
| same modified FlowNetS vs original | FlowNetS · MPI Sintel clean train | EPE: 4,50 -> 5,06 | Hurt |
| same modified FlowNetS vs original | FlowNetS · MPI Sintel Final train | EPE: 5,45 -> 5,81 | Hurt |
| add more input frames to fitted Gabor filters (ablation) | fitted c6 Gabor filter · temporal frequency bandwidth | bandwidth | Improved |
| apply ReLU and bias to the Gabor response (ablation) | fitted c6 Gabor filter · temporal frequency bandwidth | bandwidth | Improved |
Closest papers by meaning
- Models Matter, So Does Training: An Empirical Study of CNNs for Optical Flow Estimation (TPAMI 2019 · similarity 0.88)
- FlowNAS: Neural Architecture Search for Optical Flow Estimation (IJCV 2023 · similarity 0.87)
- Flow Snapshot Neurons in Action: Deep Neural Networks Generalize to Biological Motion Perception (NeurIPS 2024 · similarity 0.86)
- Neuroexplicit Diffusion Models for Inpainting of Optical Flow Fields (ICML 2024 · similarity 0.86)
- Reanimating Images using Neural Representations of Dynamic Stimuli (CVPR 2025 · similarity 0.86)
- Recurrent Partial Kernel Network for Efficient Optical Flow Estimation (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.
