Upload 234-cbot.mp4.mp4

Implemented a use case of the SO-100 arms to ignore noise. The model is trained to ignore all Lego pieces (the noise) and pickup the rest of the toys. A comparative analysis was made between all the present policies (ACT, Diffusion and smolVLA), and concluded ACT worked best for our use case, in differentiating between the noise and pick & place smoothly and swiftly.
Environmental setup: We used a monocular camera set-up (640x480p) and recorded 100 samples for training for 100,000 steps.
[LeRobot-worldwide-hackathon/234-cbot-collect-anything-not-legos](https://huggingface.co/datasets/LeRobot-worldwide-hackathon/234-cbot-collect-anything-not-legos)

234-cbot.mp4.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0adafd1433bc1a80ca5c64b22912819f1540ed96c7502ebd18c3d373c2bdfe84
+size 179101460