Research map / Video / Action recognition
DMC-Net: Generating Discriminative Motion Cues for Fast Compressed Video Action Recognition
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 flow reconstruction supervision to classification-only training (ablation) | DMC-Net · HMDB-51 | accuracy: 60.5% -> 61.5% | Helped |
| Add adversarial loss to classification and flow reconstruction losses (ablation) | DMC-Net · HMDB-51 | accuracy: 61.5% -> 61.8% | Helped |
| Replace ordinary flow reconstruction with CAM attention weighting (ablation) | DMC-Net · HMDB-51 | accuracy: 61.5 -> 61.4 | No clear effect |
| Replace ordinary flow reconstruction with GBP attention weighting (ablation) | DMC-Net · HMDB-51 | accuracy: 61.5 -> 61.5 | No clear effect |
| Add CAM attention weighting to the reconstruction loss in the GAN model (ablation) | DMC-Net with GAN · HMDB-51 | accuracy: 61.8 -> 61.5 | Hurt |
| Add GBP attention weighting to the reconstruction loss in the GAN model (ablation) | DMC-Net with GAN · HMDB-51 | accuracy: 61.8 -> 61.0 | Hurt |
| Train the generator and classifier separately rather than end-to-end (ablation) | DMC-Net · HMDB-51 | accuracy: 61.5 -> 59.3 | Hurt |
| Decompose DMC generation into two stages (ablation) | DMC-Net · HMDB-51 | accuracy: 61.5 -> 60.6 | Hurt |
| Smooth motion vectors with bilinear interpolation before generation (ablation) | DMC-Net · HMDB-51 | accuracy: 61.5 -> 61.4 | No clear effect |
| Fuse separately processed MV and residual by addition instead of direct stacking (ablation) | DMC-Net · HMDB-51 | accuracy: 61.80% -> 61.32% | Hurt |
| Fuse separately processed MV and residual by concatenation instead of direct stacking (ablation) | DMC-Net · HMDB-51 | accuracy: 61.80% -> 61.36% | Hurt |
| Use an I3D classifier instead of ResNet-18 (ablation) | DMC-Net · HMDB-51 | accuracy: 62.8 -> 71.8 | Helped |
Closest papers by meaning
- Optical Flow Guided Feature: A Fast and Robust Motion Representation for Video Action Recognition (CVPR 2018 · similarity 0.89)
- Compressed Video Action Recognition (CVPR 2018 · similarity 0.89)
- DisMo: Disentangled Motion Representations for Open-World Motion Transfer (NeurIPS 2025 · similarity 0.86)
- DVANet: Disentangling View and Action Features for Multi-View Action Recognition (AAAI 2024 · similarity 0.86)
- VA-AR: Learning Velocity-Aware Action Representations with Mixture of Window Attention (AAAI 2025 · similarity 0.86)
- TrAction: Action Recognition with Sparse Trajectories (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.