Better readme

README.md

@@ -5,13 +5,46 @@ language:
 pretty_name: human_overcooked_dataset
 ---
 
-This is a collection of human gameplay trajectories in a modified version of the game Overcooked. For more details on the environment, data collection, and methods, please refer to the paper.
+# Human Macro-Trajectory Dataset
 
-This dataset contains 225 (15 games x 15 participants) total trajectories. Each trajectory contains the following:
-- A folder of images corresponding to each frame of the game (**imgs**).
-- A **key_frames.npy**, denoting which of the images represent frames at which the team took a decision.
-- **trajectory.npy**, the micro-level trajectory, corresponding to the image frames.
-- **macro_trajectory.npy**, corresponding to the decision frames.
-- **metadata.json**, with the task metadata and agent beliefs.
+This archive contains the human trajectory data used for offline capability-inference evaluation and human-data plotting.
 
-The micro state trajectories are naturally longer than the macro states. All analysis in the paper was done using macro states.
+## Contents
+
+```text
+macro_trajectories/
+  human_macro_trajectory_dataset.npy
+  0001/
+    trajectory.npy
+    key_frames.npy
+    macro_trajectory.npy
+    metadata.json
+  0002/
+    ...
+```
+
+## File Descriptions
+human_macro_trajectory_dataset.npy
+Preprocessed dataset used directly by human_data_offline_eval.py. This is the main file needed for offline inference evaluation.
+
+trajectory.npy
+Full state trajectory recorded during human play. This contains the environment state at each low-level/micro timestep.
+
+key_frames.npy
+Indices into trajectory.npy identifying the timesteps that correspond to macro-action decision points.
+
+macro_trajectory.npy
+Subsampled trajectory containing only the states at macro-action decision points:
+```text
+macro_trajectory = trajectory[key_frames]
+```
+This is the per-trajectory format used by the diversity and trajectory-analysis plotting code.
+
+metadata.json
+Metadata for the trajectory, including the map, task, ground-truth action restrictions, timestamp, and agent-belief configuration.
+
+## Micro vs Macro States
+The Overcooked environment runs at two levels.
+Micro states are the low-level environment states observed at every primitive timestep. These correspond to individual movement/action updates, such as moving up/down/left/right or waiting while a macro action is being executed. These are stored in trajectory.npy.
+Macro states are the decision-point states where agents choose a new macro action, such as getting an ingredient, chopping, getting a plate, or delivering. These are extracted from the micro trajectory using key_frames.npy and stored in macro_trajectory.npy.
+The capability-inference algorithms operate on macro-level trajectories, so most analyses use macro_trajectory.npy or the bundled human_macro_trajectory_dataset.npy.