Upload folder using huggingface_hub

README.md

@@ -1,62 +1,114 @@
 ---
-datasets: TakuyaHiraoka/so101_pick_diverse_objects_v2
+datasets: TakuyaHiraoka/so101_pick_diverse_objects
 library_name: lerobot
 license: apache-2.0
 model_name: act
 pipeline_tag: robotics
 tags:
-- lerobot
-- act
-- robotics
+  - act
+  - so101
+  - lerobot
+  - imitation-learning
+  - human-in-the-loop
 ---
 
-# Model Card for act
+# Model Card for `act_so101_pick_diverse_objects`
 
-<!-- Provide a quick summary of what the model is/does. -->
+This is an [Action Chunking with Transformers (ACT)](https://huggingface.co/papers/2304.13705)
+policy trained to pick up a wide variety of everyday objects with the
+[SO101](https://github.com/TheRobotStudio/SO-ARM100) robotic arm.
+The policy uses **only a wrist-mounted camera** as visual input.
 
+The policy was trained with [LeRobot](https://github.com/huggingface/lerobot)
+and pushed to the Hub. See the full LeRobot documentation at
+[LeRobot Docs](https://huggingface.co/docs/lerobot/index).
 
-[Action Chunking with Transformers (ACT)](https://huggingface.co/papers/2304.13705) is an imitation-learning method that predicts short action chunks instead of single steps. It learns from teleoperated data and often achieves high success rates.
+**Demo video**
+<video controls src="https://huggingface.co/TakuyaHiraoka/act_so101_pick_diverse_objects/resolve/main/assets/act_rag.mp4"></video>
+<video controls src="https://huggingface.co/TakuyaHiraoka/act_so101_pick_diverse_objects/resolve/main/assets/act_pen.mp4"></video>
 
+---
+
+## Model Details
 
-This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
-See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).
+- **Policy class:** ACT (Action Chunking with Transformers)
+- **Robot:** SO101 (single arm)
+- **Observation:** wrist camera only (640 × 480 @ 30 fps) + joint state
+- **Framework:** LeRobot
+- **License:** apache-2.0
+- **Paper:** [arXiv:2304.13705](https://huggingface.co/papers/2304.13705)
+
+---
+
+## Training Data
+
+The policy was trained on a combination of **two datasets** collected on the
+same SO101 setup:
+
+### 1. Expert teleoperation data
+- **Dataset:** [TakuyaHiraoka/so101_pick_diverse_objects](https://huggingface.co/datasets/TakuyaHiraoka/so101_pick_diverse_objects)
+- Collected via standard teleoperation by a human expert.
+- Each episode shows the arm moving from a resting pose to grasping a target
+  object placed on a tabletop.
+- See the dataset card for details on the objects and environment.
+
+### 2. Human-in-the-Loop (HIL) correction data
+- **Dataset:** [TakuyaHiraoka/so101_pick_diverse_objects_hil_round1](https://huggingface.co/datasets/TakuyaHiraoka/so101_pick_diverse_objects_hil_round1)
+- Collected by rolling out a prototype policy (trained on the expert data
+  above) and having a human operator intervene to correct failure cases.
+
+Combining both sources is intended to improve robustness on objects and
+configurations where pure expert imitation tends to fail.
+
+---
+
+### Out-of-Scope / Limitations
+
+- The policy was trained on a specific tabletop environment and lighting
+  condition. Performance can degrade significantly under different
+  backgrounds, lighting, or camera mountings.
+- Only a wrist camera is used, so the policy has no global scene context;
+  picking objects far outside the wrist camera's field of view is unlikely
+  to work.
+- The robot embodiment is fixed (SO101 with the
+  [Innomaker U20CAM-1080P-S1](https://github.com/INNO-MAKER/U20CAM-1080P-S1)
+  wrist camera). Transferring to other arms or cameras will likely require
+  retraining or fine-tuning.
 
 ---
 
 ## How to Get Started with the Model
 
-For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
-Below is the short version on how to train and run inference/eval:
+For a complete walkthrough, see the
+[LeRobot training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
 
 ### Train from scratch
 
 ```bash
 lerobot-train \
-  --dataset.repo_id=${HF_USER}/<dataset> \
+  --dataset.repo_id=${HF_USER}/ \
   --policy.type=act \
-  --output_dir=outputs/train/<desired_policy_repo_id> \
+  --output_dir=outputs/train/ \
   --job_name=lerobot_training \
   --policy.device=cuda \
-  --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
+  --policy.repo_id=${HF_USER}/ \
   --wandb.enable=true
 ```
 
-_Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`._
+Checkpoints are written to
+`outputs/train/<desired_policy_repo_id>/checkpoints/`.
 
-### Evaluate the policy/run inference
+### Evaluate the policy / run inference
 
 ```bash
 lerobot-record \
   --robot.type=so100_follower \
-  --dataset.repo_id=<hf_user>/eval_<dataset> \
-  --policy.path=<hf_user>/<desired_policy_repo_id> \
+  --dataset.repo_id=/eval_ \
+  --policy.path=TakuyaHiraoka/act_so101_pick_diverse_objects \
   --episodes=10
 ```
 
-Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.
+Prefix the evaluation dataset repo with `eval_` and supply `--policy.path`
+pointing to this Hub checkpoint (or a local path).
 
 ---
-
-## Model Details
-
-- **License:** apache-2.0