Research map / Video / Action recognition
Guided Weak Supervision for Action Recognition with Scarce Data to Assess Skills of Children with Autism
Prashant Pandey, Prathosh AP, Manu Kohli, Josh Pritchard
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 Guided Weak Supervision and Directional Regularization | I3D · Autism dataset | accuracy: 69.3% -> 75.1% | Improved |
| Add Guided Weak Supervision and Directional Regularization | I3D · SSBD | accuracy: 91.2% -> 95.7% | Improved |
| Add GWS without Directional Regularization (ablation) | I3D · Autism dataset | accuracy: 69.3% -> 74.3% | Improved |
| Add Directional Regularization without GWS (ablation) | I3D · Autism dataset | accuracy: 69.3% -> 71.3% | Improved |
| Augment with random source samples instead of mode-matched samples (ablation) | I3D · Autism dataset | accuracy: 32% | Hurt |
| Re-train the RGB stream in addition to the optical-flow stream (ablation) | I3D+GWS+DR · Autism dataset | accuracy: 75.1% -> 68.1% | Hurt |
| Use second-best matched modes instead of best modes (ablation) | I3D+GWS+DR · Autism dataset | accuracy: 75.1% -> 74.4% | Hurt |
| Use handpicked similar source modes (ablation) | I3D+GWS+DR · Autism dataset | accuracy: 75.1% -> 75.3% | No clear effect |
| Combine best-matched modes from Kinetics and HMDB51 (ablation) | I3D+GWS+DR · Autism dataset | accuracy: 75.1% -> 74.8% | No clear effect |
| Transfer matched source modes across architectures (ablation) | TSN and I3D · Autism dataset | accuracy: 68% -> 73.1%; 69% -> 74.4% | Improved |
| Increase the amount of mode-matched source data beyond the target-data size (ablation) | Baseline I3D and TSN · Autism dataset | accuracy | Hurt |
| Use Directional Regularization with GWS during iterative augmentation (ablation) | I3D · Autism dataset | accuracy | Improved |
Closest papers by meaning
- Weakly Supervised Action Learning With RNN Based Fine-To-Coarse Modeling (CVPR 2017 · similarity 0.85)
- Large-Scale Weakly-Supervised Pre-Training for Video Action Recognition (CVPR 2019 · similarity 0.84)
- Semi-Supervised Teacher-Reference-Student Architecture for Action Quality Assessment (ECCV 2024 · similarity 0.84)
- 3D Human Sensing, Action and Emotion Recognition in Robot Assisted Therapy of Children With Autism (CVPR 2018 · similarity 0.84)
- AXON: Action Characterization Through Cross-Modal Knowledge Distillation for Neurodiverse Individuals (AAAI 2026 · similarity 0.84)
- Understanding Unique Behavioral Patterns through Multimodal Analysis of Eye-Hand Coordination in Autistic Children (Student Abstract) (AAAI 2025 · similarity 0.83)
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.
