Research map / Video / Action recognition
Long-Term Temporal Convolutions for Action Recognition
Gül Varol, Ivan Laptev, Cordelia Schmid
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 |
|---|---|---|---|
| add random clipping augmentation (ablation) | LTC 60f flow · UCF101 split 1 | video accuracy: 76.5 -> 79.6 | Improved |
| add multiscale cropping (ablation) | LTC 60f flow · UCF101 split 1 | video accuracy: 76.5 -> 78.1 | Improved |
| raise dropout to 0.9 (ablation) | LTC 60f flow · UCF101 split 1 | video accuracy: 76.5 -> 78.5 | Improved |
| combine random clipping, multiscale cropping and dropout 0.9 (ablation) | LTC 60f flow · UCF101 split 1 | video accuracy: 76.5 -> 80.5 | Improved |
| increase temporal extent from 16 to 60 frames (RGB) (ablation) | LTC RGB from scratch · UCF101 split 1 | video accuracy: 51.9 -> 59.9 | Improved |
| increase temporal extent from 16 to 60 frames (flow) (ablation) | LTC flow from scratch · UCF101 split 1 | video accuracy: 78.7 -> 80.5 | Improved |
| increase temporal extent from 16 to 60 frames (flow) (ablation) | LTC flow from scratch · HMDB51 split 1 | video accuracy: 43.9 -> 52.9 | Improved |
| pre-train flow network on UCF101 before HMDB51 fine-tuning (ablation) | LTC 60f flow · HMDB51 split 1 | video accuracy: 52.9 -> 57.1 | Improved |
| use Brox optical flow instead of RGB input (ablation) | LTC 60f from scratch · UCF101 split 1 | video accuracy: 59.9 -> 79.6 | Improved |
| extend pre-trained C3D RGB network to 100 frames (ablation) | LTC RGB · UCF101 (3 splits mean) | accuracy: 80.2% versus 82.4% | Improved |
| late fusion of flow and RGB LTC networks (ablation) | LTC · UCF101 (3 splits mean) | accuracy: 85.2 -> 91.7 | Improved |
| add IDT features to LTC Flow+RGB (ablation) | LTC Flow+RGB · HMDB51 (3 splits mean) | accuracy: 64.8 -> 67.2 | Improved |
Closest papers by meaning
- Memory-Augmented Temporal Dynamic Learning for Action Recognition (AAAI 2019 · similarity 0.91)
- Temporal Convolutional Networks for Action Segmentation and Detection (CVPR 2017 · similarity 0.91)
- Deep Analysis of CNN-Based Spatio-Temporal Representations for Action Recognition (CVPR 2021 · similarity 0.90)
- Deep Temporal Linear Encoding Networks (CVPR 2017 · similarity 0.90)
- Lattice Long Short-Term Memory for Human Action Recognition (ICCV 2017 · similarity 0.90)
- T-C3D: Temporal Convolutional 3D Network for Real-Time Action Recognition (AAAI 2018 · similarity 0.90)
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.
