cut_act_thesis_cam3_omly
robotics policy for precision manipulation
1 min readupdated
This is a ready-to-use robotics policy trained with ACT (Action Chunking with Transformers, an imitation learning method that predicts short action sequences instead of single steps). Built for cutting tasks on a robotic arm with a specific camera setup and joint-state configuration, it was trained with the LeRobot library on teleoperation demonstrations. It is well suited for researchers and developers working in home or lab robotics.
Key features
- Robot type: compatible with the so101_follower robotic arm.
- Cameras: uses feeds from two cameras (wrist and side).
- Architecture: built on Action Chunking with Transformers (ACT).
- Dataset: trained on 103 episodes (72,725 frames at 30 FPS).
What you can do
- Run the pre-trained policy on your robot via the LeRobot CLI utility.
- Use the task demonstration dataset to fine-tune your own models.
- Tune inference parameters to automate object manipulation.
Example
lerobot-rollout \
--strategy.type=base \
--robot.type=so101_follower \
--robot.port= \
--robot.cameras="{ : {type: opencv, index_or_path: , width: 640, height: 480, fps: 30}, : {type: opencv, index_or_path: , width: 640, height: 480, fps: 30}}" \
--policy.path=Kimura-881014/cut_act_thesis_cam3_omly \
--task="cut entangle chips" \
--duration=60Read more
DetailsTraining steps: 20,000
Batch size: 8
Optimizer: adamw (Learning rate: 1e-05)
LeRobot version: 0.7.1










Comments