Research map / Probabilistic and scientific modeling / Molecules and proteins
Bidirectional Progressive Transformer for Interaction Intention Anticipation
Zichen Zhang, Hongchen Luo, Wei Zhai, Yu Kang, Yang Cao
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 |
|---|---|---|---|
| alternate cross-attention refinement between trajectory and hotspot features at each time step (Bi-Progressive) instead of separate prediction (ablation) | BOT · joint-forecasting ablation | NSS: 0.86 -> 1.70 | Improved |
| add progressive time-step alternation on top of one-shot bidirectional conditioning (ablation) | BOT · joint-forecasting ablation | ADE: 0.10 -> 0.09 | Improved |
| condition hotspots on hand trajectory only (ablation) | BOT · joint-forecasting ablation | SIM: 0.18 -> 0.25 | Improved |
| use heatmap rather than point as the C-VAE condition (ablation) | BOT hotspot decoder · ablation | hotspot accuracy | Improved |
| add cross-attention fusion of consecutive trajectory points (condition vs none) (ablation) | BOT trajectory branch · ablation | trajectory accuracy | Improved |
| increase trajectory uncertainty scale alpha beyond a moderate value (ablation) | Trajectory Stochastic Unit · uncertainty analysis | ADE/FDE (20 samples) | Mixed |
| use stochastic unit instead of repeated full inference for multi-sample trajectories | BOT · inference time | inference time (20 samples): 923ms (OCT*) -> 92ms | Improved |
| use joint BOT instead of best non-joint video/image model (Segformer-b2), single sample on EGO4D | interaction intention anticipation · EGO4D | ADE: 0.16 -> 0.11 | Improved |
Closest papers by meaning
- AFF-ttention! Affordances and Attention models for Short-Term Object Interaction Anticipation (ECCV 2024 · similarity 0.89)
- Forecasting Human-Object Interaction: Joint Prediction of Motor Attention and Actions in First Person Video (ECCV 2020 · similarity 0.89)
- Joint Hand Motion and Interaction Hotspots Prediction From Egocentric Videos (CVPR 2022 · similarity 0.89)
- How Do I Do That? Synthesizing 3D Hand Motion and Contacts for Everyday Interactions (CVPR 2025 · similarity 0.86)
- Summarize the Past to Predict the Future: Natural Language Descriptions of Context Boost Multimodal Object Interaction Anticipation (CVPR 2024 · similarity 0.85)
- Social-Transmotion: Promptable Human Trajectory Prediction (ICLR 2024 · similarity 0.85)
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.
