Preprocessed dataset for my paper "You Only Teach Once: Learn One-Shot Bimanual Robotic Manipulation from Video Demonstrations" [arXiv] / [Project] / [Code]
Please refer AugDemos and BiDP for the usage of these uploaded datasets and pretrained models.
Raw left/right RGB images and dual-arm robot actions: drawer.zip, pouring.zip, unscrew.zip, uncover.zip, openbox.zip.
RGB images of segmented manipulated objects: drawer_masks.zip, pouring_masks.zip, unscrew_masks.zip, uncover_masks.zip, openbox_masks.zip.
Preprocessed datasets without augmentation: drawer_preprocessed.json, pouring_preprocessed.json, unscrew_preprocessed.json, uncover_preprocessed.json, openbox_preprocessed.json.
Preprocessed datasets with 100x augmentation: drawer_preprocessed_aug100x_1~4.json, pouring_preprocessed_aug100x_1~3.json, unscrew_preprocessed_aug100x.json, uncover_preprocessed_aug100x.json, openbox_preprocessed_aug100x.json.
Pretrained noaug BiDP models: bidp_drawer_noaug_ckpt01999.pth, bidp_pouring_noaug_ckpt01999.pth, bidp_unscrew_noaug_ckpt03999.pth, bidp_uncover_noaug_ckpt03999.pth, bidp_openbox_noaug_ckpt03999.pth.
Pretrained withaug BiDP models: bidp_drawer_withaug_ckpt00499.pth, bidp_pouring_withaug_ckpt00499.pth, bidp_unscrew_withaug_ckpt00999.pth, bidp_uncover_withaug_ckpt00999.pth, bidp_openbox_withaug_ckpt00999.pth.