Research map / Reinforcement learning / Visual and representation RL
First-Person Activity Forecasting With Online Inverse Reinforcement Learning
Nicholas Rhinehart, Kris M. Kitani
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 stop-based instead of scene-based goal discovery (ablation) | Darko · Goal forecasting, labelled actions, Home 1 | mean true-goal probability: 0.438 -> 0.614 | Improved |
| use stop-based instead of scene-based goal discovery (ablation) | Darko · Goal forecasting, labelled actions, Office 2 | mean true-goal probability: 0.238 -> 0.625 | Improved |
| use stop-based instead of scene-based goal discovery (ablation) | Darko · Goal forecasting, labelled actions, Lab 1 | mean true-goal probability: 0.426 -> 0.709 | Improved |
| use ground-truth labelled detections instead of visual detections (ablation) | Darko · Goal forecasting, Home 1 | mean true-goal probability: 0.524 -> 0.851 | Improved |
| add action features to state features in the reward (ablation) | Darko with state-only features · Goal forecasting, labelled detections, Home 1 | mean true-goal probability: 0.735 -> 0.851 | Improved |
| add state features to position-only reward features (ablation) | Darko with position-only features · Goal forecasting, labelled detections, Home 1 | mean true-goal probability: 0.674 -> 0.851 | Improved |
| use full state-action reward features instead of state only (ablation) | Darko · Goal forecasting, labelled detections, Lab 1 | mean true-goal probability: 0.892 -> 0.880 | No clear effect |
| add goal-detection uncertainty (log-probability goal values) (ablation) | Darko · Synthetic noisy goal detections | mean true-goal probability | Improved |
| use Darko online IRL instead of logistic regression | Logistic · Goal forecasting, visual detections, Home 1 | mean true-goal probability: 0.458 -> 0.524 | Improved |
| use Darko online IRL instead of RNN | RNN · Goal forecasting, visual detections, Lab 1 | mean true-goal probability: 0.455 -> 0.473 | Improved |
| use Darko trajectory length forecast instead of nearest neighbors | Nearest neighbors · Trajectory length, Office 1 | median % error: 42.9 -> 17.3 | Improved |
| use Darko trajectory length forecast instead of nearest neighbors | Nearest neighbors · Trajectory length, Home 1 | median % error: 29.0 vs 30.0 | No clear effect |
Closest papers by meaning
- First-Person Activity Forecasting from Video with Online Inverse Reinforcement Learning (TPAMI 2018 · similarity 0.99)
- Visual Forecasting by Imitating Dynamics in Natural Sequences (ICCV 2017 · similarity 0.88)
- Predicting Goal-Directed Human Attention Using Inverse Reinforcement Learning (CVPR 2020 · similarity 0.88)
- Generative Hybrid Representations for Activity Forecasting With No-Regret Learning (CVPR 2020 · similarity 0.87)
- Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning (NeurIPS 2024 · similarity 0.85)
- Grounding Video Models to Actions through Goal Conditioned Exploration (ICLR 2025 · 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.
