Research map / Video / Action recognition
Spatiotemporal Fusion in 3D CNNs: A Probabilistic View
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 |
|---|---|---|---|
| Select an optimized fusion strategy from the learned probability space rather than using a fixed fusion unit throughout (ablation) | 3D DenseNet121 · SomethingV1 | Accuracy: 41.8% -> 50.2% | Helped |
| Select an optimized fusion strategy from the learned probability space rather than using a fixed fusion unit throughout (ablation) | 3D DenseNet121 · SomethingV2 | Accuracy: 55.1% -> 62.4% | Helped |
| Select an optimized fusion strategy from the learned probability space rather than using a fixed fusion unit throughout (ablation) | 3D DenseNet121 · UCF101 | Accuracy: 84.2% -> 84.2% | No clear effect |
| Select an optimized fusion strategy from the learned probability space rather than using a fixed fusion unit throughout (ablation) | 3D DenseNet121 · Kinetics400 | Accuracy: 69.7% -> 71.7% | Helped |
| Use S+ST in the first half and last three layers, and ST in the remaining layers (ablation) | 3D ResNet50 · Something V1 (clip-level) | Accuracy: 40.1% -> 41.2% | Helped |
| Use S+ST in the first half and last three layers, and ST in the remaining layers (ablation) | 3D ResNeXt50 · Something V1 (clip-level) | Accuracy: 42.1% -> 43.6% | Helped |
| Use S+ST in the first half and last three layers, and ST in the remaining layers (ablation) | 3D ResNeXt101 · Something V1 (clip-level) | Accuracy: 42.7% -> 44.0% | Helped |
| Use the probability-space-selected fusion strategy | 3D DenseNet121 · Something-Something V1 | Top-1 accuracy: 49.8% -> 50.2% | Helped |
| Use the probability-space-selected fusion strategy | 3D DenseNet121 · Something-Something V2 | Val. Top-1: 59.1% -> 62.9% | Helped |
| Use the probability-space-selected fusion strategy | 3D DenseNet121 · Kinetics400 | Top1 accuracy: 72.0% -> 72.5% | Helped |
| Use the proposed fusion strategy with ImageNet pretraining | UCF101 · UCF101 (ImageNet pretraining) | Top-1 accuracy: 84.5% -> 86.0% | Helped |
| Use the proposed fusion strategy with Kinetics400 pretraining | UCF101 · UCF101 (Kinetics400 pretraining) | Top-1 accuracy: 96.8% -> 96.5% | Hurt |
Closest papers by meaning
- Spatiotemporal Pyramid Network for Video Action Recognition (CVPR 2017 · similarity 0.92)
- A Closer Look at Spatiotemporal Convolutions for Action Recognition (CVPR 2018 · similarity 0.91)
- Deep Analysis of CNN-Based Spatio-Temporal Representations for Action Recognition (CVPR 2021 · similarity 0.90)
- Collaborative Spatiotemporal Feature Learning for Video Action Recognition (CVPR 2019 · similarity 0.90)
- Convolutional Two-Stream Network Fusion for Video Action Recognition (CVPR 2016 · similarity 0.89)
- What Have We Learned From Deep Representations for Action Recognition? (CVPR 2018 · similarity 0.89)
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.