dp-cupbasket-visual-8k

LeRobot diffusion policy (ROBOTIS lerobot-cyclo fork, 288 M parameters) for the bimanual cup handover onto a basket with the ROBOTIS FFW SG2 (SG2 + CAD grippers, calibrated wrist cameras), visual DR, wide + narrow placement (~8.6k demos).

Files at the repo root = step 100k (DiffusionPolicy.from_pretrained("padiredd/dp-cupbasket-visual-8k")); step-10000/, step-20000/, step-30000/, step-40000/, step-50000/, step-60000/, step-70000/, step-80000/, step-90000/ = intermediate checkpoints (every 10k steps).

data padiredd/Cup_visual_8k: all ~6.6k successes of the wide-placement visual-DR cup collection (data_collection/cup_basket/visual10k_wide, 2026-11-08/09; the cup DR ablation's VISUAL recipe: v6 program cup_handover_v6.py, right close recorded 0.70 / executed 0.50, cup-safe gripper randomization, table look / backdrop / cup colour / room + spot light randomization; placement widened to cup x 0.36-0.58 y 0.00-0.42 m, basket x 0.40-0.55 y -0.60..-0.28 m, yaw +-10 deg; seeds 100000+). The 10k collection was stopped by the user at ~6.6k successes (padiredd/Cup_visual_wide_1..4) followed by the 2,000 demos of padiredd/Ablation_Cup_visual (same visual recipe, narrower placement: cup x 0.44-0.58 y 0.10-0.30, basket x 0.40-0.54 y -0.52..-0.36)
sampling uniform over all frames (LeRobot default EpisodeAwareSampler)
inputs / outputs head + left wrist + right wrist cameras, each 240x240, + 22-dim joint state; action = 16 joint targets, 15 Hz
image augmentation heavy, random order, up to 5 of: brightness 0.6-1.4, contrast 0.6-1.4, saturation 0.3-1.7, hue +-0.1, sharpness 0.3-2.0, affine +-10 deg / 10 % / 0.9-1.1, perspective 0.2, blur, random erasing, grayscale
state noise none
recipe 100k steps, batch 32, ResNet18 per camera, Adam 1e-4, cosine, DDPM 100 steps, horizon 64 / 32 executed, n_obs_steps 1, seed 1000
hardware one RTX 4090 24 GB (lab225), shared with dp-cupbasket-visual-wide-500 (~4.9 steps/s each)

Task string: "pick up the blue cup with the left hand, hand it to the right hand, and place it on the basket". Evaluation results are added once they are run.

Downloads last month
37
Safetensors
Model size
0.3B params
Tensor type
F32
·
Video Preview
loading

Collection including padiredd/dp-cupbasket-visual-8k