Research map / Video / Action recognition
SpeedNet: Learning the Speediness in Videos
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 |
|---|---|---|---|
| Add same-batch pairing of normal-speed and sped-up versions of each clip (ablation) | SpeedNet · NFS | accuracy: 59.3% -> 73.6% | Helped |
| Add temporal augmentation (ablation) | SpeedNet without same-batch pairing · NFS | accuracy: 57.7% -> 59.3% | Helped |
| Add spatial augmentation (ablation) | SpeedNet without same-batch pairing or temporal augmentation · NFS | accuracy: 57.4% -> 57.7% | Helped |
| Add same-batch pairing and temporal and spatial augmentations (ablation) | SpeedNet · Kinetics test | accuracy: 96.9% -> 75.6% | Hurt |
| Add same-batch pairing and temporal and spatial augmentations (ablation) | SpeedNet · NFS | accuracy: 57.4% -> 73.6% | Helped |
| Use SpeedNet self-supervised pre-training before fine-tuning | S3D-G action recognition · UCF101 split-1 | accuracy: 73.8 -> 81.1 | Helped |
| Use SpeedNet self-supervised pre-training before fine-tuning | S3D-G action recognition · HMDB51 split-1 | accuracy: 46.4 -> 48.8 | Helped |
| Use SpeedNet self-supervised pre-training before fine-tuning | I3D action recognition · UCF101 split-1 | accuracy: 47.9 -> 66.7 | Helped |
| Use SpeedNet self-supervised pre-training before fine-tuning | I3D action recognition · HMDB51 split-1 | accuracy: 29.6 -> 43.7 | Helped |
Closest papers by meaning
- RSPNet: Relative Speed Perception for Unsupervised Video Representation Learning (AAAI 2021 · similarity 0.90)
- FASTER Recurrent Networks for Efficient Video Classification (AAAI 2020 · similarity 0.89)
- Semi-Supervised Action Recognition With Temporal Contrastive Learning (CVPR 2021 · similarity 0.88)
- Learning Motion Patterns in Videos (CVPR 2017 · similarity 0.88)
- SlowFast-VGen: Slow-Fast Learning for Action-Driven Long Video Generation (ICLR 2025 · similarity 0.86)
- TTVidT: Decoupling the Temporal Axis for Efficient Motion-Centric Video Pretraining (NeurIPS 2026 · 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.