Research map / Video / Action recognition
Deep Temporal Linear Encoding Networks
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 |
|---|---|---|---|
| aggregate temporal feature maps by element-wise multiplication rather than averaging or maximum (ablation) | BN-Inception two-stream TLE · UCF101 and HMDB51 (split1) | accuracy: 91.3/67.4; 92.6/68.1 -> 94.8/70.4 | Helped |
| aggregate C3D clip features by element-wise multiplication rather than averaging or maximum (ablation) | C3D TLE · UCF101 (split1) | accuracy: 84.2; 84.6 -> 86.1 | Helped |
| use BN-Inception instead of VGG-16 or AlexNet (ablation) | two-stream TLE · UCF101 and HMDB51 (split1), spatial and temporal streams | accuracy: AlexNet 74.4/50.8, 82.7/52.4; VGG-16 81.5/60.9, 86.8/61.5; BN-Inception 86.9/63.2, 89.1/66.4 | Helped |
| use bilinear encoding rather than fully connected pooling (ablation) | BN-Inception two-stream TLE · UCF101 and HMDB51 | accuracy: 92.2/68.8 -> 95.6/71.1 | Helped |
| use full bilinear encoding rather than Tensor Sketch bilinear encoding (ablation) | BN-Inception two-stream TLE · UCF101 and HMDB51 | accuracy: 95.1/70.6 -> 95.6/71.1 | Helped |
| use bilinear encoding rather than fully connected pooling (ablation) | C3D TLE · UCF101 and HMDB51 | accuracy: 83.1/58.6 -> 86.3/60.3 | Helped |
| use full bilinear encoding rather than Tensor Sketch bilinear encoding (ablation) | C3D TLE · UCF101 and HMDB51 | accuracy: 85.6/59.7 -> 86.3/60.3 | Helped |
| add a Places365-pretrained scene-context stream (ablation) | VGG-16 spatial TLE · UCF101 and HMDB51 (split1) | accuracy: 81.5 / 60.9 -> 83.8 / 63.6 | Helped |
| encode entire videos with TLE bilinear features rather than the Two Stream baseline | BN-Inception two-stream · UCF101 and HMDB51 | accuracy: 88.0/59.4 -> 95.6/71.1 | Helped |
| add TLE bilinear encoding to the C3D model | C3D · UCF101 and HMDB51 | accuracy: 82.3/56.8 -> 86.3/60.3 | Helped |
Closest papers by meaning
- T-C3D: Temporal Convolutional 3D Network for Real-Time Action Recognition (AAAI 2018 · similarity 0.89)
- Deep Analysis of CNN-Based Spatio-Temporal Representations for Action Recognition (CVPR 2021 · similarity 0.88)
- Temporal Convolutional Networks for Action Segmentation and Detection (CVPR 2017 · similarity 0.88)
- TDN: Temporal Difference Networks for Efficient Action Recognition (CVPR 2021 · similarity 0.87)
- Masked Motion Encoding for Self-Supervised Video Representation Learning (CVPR 2023 · similarity 0.86)
- Enhancing Temporal Understanding in Video-LLMs through Stacked Temporal Attention in Vision Encoders (NeurIPS 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.