Research map / Video / Action recognition
Sympathy for the Details: Dense Trajectories and Hybrid Classification Architectures for Action Recognition
César Roberto de Souza, Adrien Gaidon, Eleonora Vig, Antonio Manuel López
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 data augmentation by feature stacking (DAFS) (ablation) | iDT+STA+MIFS+DN · UCF-101 | mean accuracy (%): 89.2 -> 90.6 | Improved |
| add data augmentation by feature stacking (DAFS) (ablation) | iDT+STA+MIFS+DN · HMDB-51 | mean accuracy (%): 65.4 -> 67.8 | Improved |
| add double normalization (ablation) | iDT · HMDB-51 | mean accuracy (%): 57.0 -> 59.1 | Improved |
| add spatio-temporal augmentation (STA) (ablation) | iDT · HMDB-51 | mean accuracy (%): 57.0 -> 60.3 | Improved |
| add fine-tuned C and Gaussian kernel to SVM (ablation) | FV-SVM · five datasets | accuracy | No clear effect |
| add a supervised deep network after unsupervised PCA reduction (depth 1) | best FV-SVM · five datasets (average) | relative improvement: 1.28% | Improved |
| use two layers of width 4096 batch 128 after PCA (ablation) | best FV-SVM · five datasets (average) | relative improvement: 2.46% | Improved |
| learn dimensionality reduction layer supervised instead of PCA (ablation) | hybrid depth 1 · UCF-101 | mean accuracy (%): 91.9 -> 92.3 | Improved |
| learn dimensionality reduction layer supervised instead of PCA (ablation) | hybrid depth 1 · Hollywood2 | mAP (%): 73.0 -> 72.5 | Hurt |
| transfer PCA weights from UCF-101 (ablation) | hybrid · High-Five | mAP+ (%): 73.1 -> 76.3 | Improved |
| transfer PCA weights from UCF-101 (ablation) | hybrid · HMDB-51 | mean accuracy (%): 68.0 -> 66.5 | Hurt |
| transfer GMMs from UCF-101 (ablation) | hybrid · Hollywood2 | mAP (%): 72.6 -> 72.4 | No clear effect |
Closest papers by meaning
- Deep Analysis of CNN-Based Spatio-Temporal Representations for Action Recognition (CVPR 2021 · similarity 0.91)
- No Frame Left Behind: Full Video Action Recognition (CVPR 2021 · similarity 0.90)
- Unsupervised Deep Learning of Mid-Level Video Representation for Action Recognition (AAAI 2018 · similarity 0.90)
- Adversarial Action Prediction Networks (TPAMI 2018 · similarity 0.90)
- Action Recognition With Spatial-Temporal Discriminative Filter Banks (ICCV 2019 · similarity 0.90)
- Discriminative Video Representation Learning Using Support Vector Classifiers (TPAMI 2019 · 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.
