Research map / Video / Action recognition
Slow and Steady Feature Analysis: Higher Order Temporal Coherence in Video
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 second-order temporal steadiness with first-order slowness instead of first-order SFA alone | SSFA · NORB→NORB | accuracy (%): 39.23±0.94 -> 42.83±0.33 | Helped |
| Use second-order temporal steadiness with first-order slowness instead of first-order SFA alone | SSFA · KITTI→SUN | accuracy (%): 1.21±0.14 -> 1.65±0.04 | Helped |
| Use second-order temporal steadiness with first-order slowness instead of first-order SFA alone | SSFA · KITTI→SUN top-10 | accuracy (%): 8.24±0.25 -> 9.19±0.10 | Helped |
| Use second-order temporal steadiness with first-order slowness instead of first-order SFA alone | SSFA · HMDB→PASCAL-10 | accuracy (%): 19.26±0.45 -> 20.95±0.13 | Helped |
| Replace HMDB unlabeled video with KITTI video (ablation) | SSFA · PASCAL-10 action recognition | accuracy (%): 20.95% -> 20.01% | Hurt |
| Increase SUN labeled training examples from 6 to 20 per class (ablation) | SSFA · KITTI→SUN | accuracy (%): 2.65% -> 3.24% | Helped |
| Increase the amount of HMDB unlabeled video used for training (ablation) | SSFA · HMDB→PASCAL-10 | accuracy (%): 18.06, 19.74, 20.36, and 20.95% | Helped |
| Train SSFA using its unsupervised loss alone rather than jointly with a supervised objective (ablation) | SSFA features · PASCAL-10 k-NN classification | accuracy (%): ≈ 17.8% -> 19.62, 20.32 and 22.14% | Helped |
| Use SSFA unsupervised video pretraining instead of supervised CIFAR-100 pretraining | PASCAL-10 action recognition · PASCAL-10 | accuracy | Helped |
| Use SSFA unsupervised video pretraining instead of supervised CIFAR-100 pretraining | SUN scene recognition · SUN | accuracy | Mixed |
Closest papers by meaning
- Single-Frame Regularization for Temporally Stable CNNs (CVPR 2019 · similarity 0.87)
- Smooth Regularization for Efficient Video Recognition (NeurIPS 2025 · similarity 0.87)
- DejaVid: Encoder-Agnostic Learned Temporal Matching for Video Classification (CVPR 2025 · similarity 0.86)
- Temporally Consistent Object-Centric Learning by Contrasting Slots (CVPR 2025 · similarity 0.86)
- Chirality in Action: Time-Aware Video Representation Learning by Latent Straightening (NeurIPS 2025 · similarity 0.86)
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 2025 · similarity 0.86)
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.