Research map / 3D vision / 3D generation
ObjectForesight: Predicting 3D Object Trajectories from Human Videos
Rustin Soraki, Homanga Bharadhwaj, Ali Farhadi, Roozbeh Mottaghi
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 diffusion transformer instead of autoregressive transformer (ablation) | ObjectForesight · Epic-Kitchens | ARE: 9.48° -> 2.30° | Improved |
| use diffusion transformer instead of autoregressive transformer (ablation) | ObjectForesight · HOT3D-Clips | ADE: 0.055 -> 0.021 | Improved |
| use ObjectForesight-DiT instead of Constant Velocity | ObjectForesight · Epic-Kitchens | ADE: 0.027 -> 0.016 | Improved |
| predict motion directly in SE(3) instead of recovering poses from Luma Ray3 video generation | ObjectForesight-DiT · Epic-Kitchens (20-video subset) | ADE: 0.084 -> 0.029 | Improved |
| add PTV3 scene encoder instead of no encoder (ablation) | ObjectForesight-DiT · Epic-Kitchens | ARE: 2.690° -> 2.299° | Improved |
| use SparseConv scene encoder instead of no encoder (ablation) | ObjectForesight-DiT · Epic-Kitchens | ARE: 2.690° -> 2.700° | Hurt |
| scale DiT from 6L-384D to 12L-768D (ablation) | ObjectForesight-DiT · Epic-Kitchens | ARE: 4.242° -> 2.299° | Improved |
| use 3 context frames instead of 1 (ablation) | ObjectForesight-DiT · Epic-Kitchens | ARE: 7.97° -> 2.30° | Improved |
| increase context length from 3 to 10 frames (ablation) | ObjectForesight-DiT · Epic-Kitchens | ARE: 2.30° -> 8.09° | Hurt |
| train with horizon H=8 instead of H=4, evaluated at H=4 (ablation) | ObjectForesight-DiT · Epic-Kitchens | ADE: 0.0161 -> 0.0098 | Improved |
| train with horizon H=32 instead of H=8, evaluated at short horizons (ablation) | ObjectForesight-DiT · Epic-Kitchens | ADE: H=8: 0.016 -> H=32: 0.023 at H=8 eval; long-horizon ARE at H=16 5.11° -> 4.94° | Mixed |
Closest papers by meaning
- OCK: Unsupervised Dynamic Video Prediction with Object-Centric Kinematics (ICCV 2025 · similarity 0.89)
- Gaze Beyond the Frame: Forecasting Egocentric 3D Visual Span (NeurIPS 2025 · similarity 0.88)
- MolmoMotion: Forecasting Point Trajectories in 3D with Language Instruction (NeurIPS 2026 · similarity 0.88)
- Vision-Guided Action: Enhancing 3D Human Motion Prediction with Gaze-informed Affordance in 3D Scenes (CVPR 2025 · similarity 0.88)
- PlaySlot: Learning Inverse Latent Dynamics for Controllable Object-Centric Video Prediction and Planning (ICML 2025 · similarity 0.87)
- DreamScene4D: Dynamic Multi-Object Scene Generation from Monocular Videos (NeurIPS 2024 · similarity 0.87)
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.
