---
license: apache-2.0
base_model: lerobot/act
tags:
  - lerobot
  - act
  - robotics
  - manipulation
  - real-robot
  - so101
  - visuomotor
datasets:
  - ShubhamK32/so101_declutter_v1
pipeline_tag: robotics
---

# ACT — SO-101 Space Decluttering

ACT (Action Chunking Transformer) policy trained on the [SO-101 Space Decluttering Dataset v1](https://huggingface.co/datasets/ShubhamK32/so101_declutter_v1) for pick-and-place decluttering tasks on a 6-DoF SO-101 robotic arm. Trained using [LeRobot](https://github.com/huggingface/lerobot).

## Training Details

- **Policy:** ACT (Action Chunking Transformer)
- **Steps:** 100,000
- **Robot:** SO-101 6-DoF leader-follower
- **Cameras:** Dual-view — fixed top-view + wrist-mounted egocentric
- **Framework:** LeRobot

## Dataset

Trained on [ShubhamK32/so101_declutter_v1](https://huggingface.co/datasets/ShubhamK32/so101_declutter_v1) — a multi-view teleoperation dataset with spatial distractors injected to prevent visual shortcut learning.

## Usage
```python
from lerobot.policies.act.modeling_act import ACTPolicy

policy = ACTPolicy.from_pretrained("ShubhamK32/act_so101_declutter")
```

## Camera Views

- `observation.images.topview` — Fixed overhead. Better for unoccluded pick-place tasks.
- `observation.images.wristview` — Egocentric wrist-mounted. Better for overlapping and cluttered scenes.

## Related

- Dataset: [ShubhamK32/so101_declutter_v1](https://huggingface.co/datasets/ShubhamK32/so101_declutter_v1)
- SmolVLA checkpoint: [ShubhamK32/smolvla_so101_declutter](https://huggingface.co/ShubhamK32/smolvla_so101_declutter)