RiddhiCh/LGRM_BridgeDataV2_Teleoperated

SGE-Goal BridgeDataV2 (teleoperated)

Training and evaluation data for SGE-Goal, from Goal Images Are Control Targets: Spatially-Grounded Goal-Image Synthesis for Robotic Manipulation (WACV 2027).

• Code: https://github.com/Riddhi-Chatterjee/SGE-Goal
• Project page: https://riddhi-chatterjee.github.io/SGE-Goal/
• Model checkpoints: https://huggingface.co/RiddhiCh/SGE-Goal

Each sample turns one atomic manipulation step from a robot episode into an image-editing example: the scene before the step, the scene after it, where the edit happens, and the instruction. The samples were extracted automatically from BridgeData V2, teleoperated trajectories, without manual annotation.

Fields

Field	Type	Description
id	string	Sample identifier: the source trajectory path, e.g. bridge_data_v2_datacol2_toykitchen2_many_skills_34_2023-03-09_12-34-29_raw_traj_group0_traj9_images0
original_image	image	Observation before the manipulation step (I_o)
edited_image	image	Observation after the step, the goal image (I_e)
initial_edit_region_mask	image	Binary mask of the manipulated objects before the step (M_i)
final_edit_region_mask	image	Binary mask of where those objects are after the step (M_f)
edit_instruction	string	The atomic manipulation instruction (T)

Images are 640 × 480 pixels; masks are white (edit region) on black.

About half of the instructions keep the annotation-confidence line from BridgeData V2's language files (for example open the silver pot …\nconfidence: 0.2). The released SGE-Goal models were trained on the field as-is; strip the line if you need plain instructions.

Splits

Split	Samples
train	40,851
validation	1,000
test	2,000

Samples are assigned to splits by MD5 hashing of their identifier (seed 42).

Usage

from datasets import load_dataset

ds = load_dataset("RiddhiCh/LGRM_BridgeDataV2_Teleoperated")
sample = ds["train"][0]
sample["original_image"]      # PIL.Image
sample["edit_instruction"]    # str

How the data was extracted

The SGE-Goal data pipeline chains open vision models on each episode:

1. robot-arm masking and reference-frame selection (EVF-SAM, verified with BLIP-2),
2. class-agnostic object proposals on the reference frame (SAM),
3. object tracking through the episode (XMem),
4. separation of moved and stationary objects with LoFTR matching and 2-means clustering, followed by edit-mask generation and selection of the edited frame.

For episodes without a language instruction, VideoChat-Flash writes one from the video. The pipeline is in data_pipeline/. In a study with 50 annotators rating 500 tuples per dataset (1–5 scale), the lowest mean rating of any extracted component across the three datasets was 4.44.

Intended use

Training and evaluating goal-image generators and language-guided image editing for robotic manipulation. Masks and edited frames come from an automatic pipeline: they are reliable in aggregate, but individual samples can be imperfect, typically with heavily occluded targets or several near-identical objects.

Licence

BridgeData V2 is released under the Creative Commons Attribution 4.0 International License; this derived dataset is released under the same licence.

Citation

If you use this dataset, please cite SGE-Goal and the source dataset. The SGE-Goal entry is temporary until the proceedings are published.

@inproceedings{chatterjee2027goalimages,
  title     = {Goal Images Are Control Targets: Spatially-Grounded Goal-Image Synthesis for Robotic Manipulation},
  author    = {Chatterjee, Riddhi and Shrirao, Ritish and Mehrotra, Kushagra and Gopalakrishnan, Viswanath},
  booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
  year      = {2027}
}

@inproceedings{walke2023bridgedata,
  title     = {BridgeData V2: A Dataset for Robot Learning at Scale},
  author    = {Walke, Homer and Black, Kevin and Lee, Abraham and Kim, Moo Jin and Du, Max and Zheng, Chongyi and Zhao, Tony and Hansen-Estruch, Philippe and Vuong, Quan and He, Andre and Myers, Vivek and Fang, Kuan and Finn, Chelsea and Levine, Sergey},
  booktitle = {Conference on Robot Learning (CoRL)},
  year      = {2023}
}