Research map / Video / Video-language understanding
Location-Aware Graph Convolutional Networks for Video Question Answering
Deng Huang, Peihao Chen, Runhao Zeng, Qing Du, Mingkui Tan, Chuang Gan
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 |
|---|---|---|---|
| use a location-aware object graph with GCN reasoning | Video QA · TGIF-QA | accuracy: 70.58 -> 74.32 | Improved |
| add detected object features (ablation) | global-frame baseline · TGIF-QA | accuracy: 70.58 -> 72.82 | Improved |
| add GCN object-relation reasoning (ablation) | object-feature variant · TGIF-QA | accuracy: 72.82 -> 74.10 | Improved |
| replace GCNs with two FC layers (ablation) | object-feature variant · TGIF-QA | accuracy: 74.10 -> 72.96 | Hurt |
| replace GCNs with a 2-layer LSTM (ablation) | object-feature variant · TGIF-QA | accuracy: 74.10 -> 72.65 | Hurt |
| add spatial and temporal location features (ablation) | object-feature GCN variant · TGIF-QA | accuracy: 74.10 -> 74.32 | Improved |
| use temporal location only instead of both location types (ablation) | L-GCN · TGIF-QA | accuracy: 74.32 -> 73.75 | Hurt |
| use spatial location only instead of both location types (ablation) | L-GCN · TGIF-QA | accuracy: 74.32 -> 73.58 | Hurt |
| increase GCN depth from 2 to 3 layers (ablation) | L-GCN · TGIF-QA Action | accuracy: 74.32 -> 74.32 | No clear effect |
| increase GCN depth from 2 to 4 layers (ablation) | L-GCN · TGIF-QA Action | accuracy: 74.32 -> 73.97 | Hurt |
| use 2 objects per frame instead of 5 (ablation) | L-GCN · TGIF-QA Action | accuracy: 74.32 -> 74.11 | Hurt |
| use 10 objects per frame instead of 5 (ablation) | L-GCN · TGIF-QA Action | accuracy: 74.32 -> 74.01 | Hurt |
Closest papers by meaning
- (2.5+1)D Spatio-Temporal Scene Graphs for Video Question Answering (AAAI 2022 · similarity 0.92)
- Divide and Conquer: Question-Guided Spatio-Temporal Contextual Attention for Video Question Answering (AAAI 2020 · similarity 0.90)
- Learning Situation Hyper-Graphs for Video Question Answering (CVPR 2023 · similarity 0.90)
- Video Graph Transformer for Video Question Answering (ECCV 2022 · similarity 0.89)
- Tool-Augmented Spatiotemporal Reasoning for Streamlining Video Question Answering Task (NeurIPS 2025 · similarity 0.88)
- STAIR: Spatial-Temporal Reasoning with Auditable Intermediate Results for Video Question Answering (AAAI 2024 · similarity 0.88)
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.
