The robotics data ecosystem keeps getting stronger 🦾
The author explores the LeRobot community dataset, using embeddings and FiftyOne to analyze 497 episodes across 50 robot types. They identify near-duplicates, filter sessions, and reveal cross-embodiment patterns like cloth folding. A curated subset is available for further exploration.
The author demonstrates how to analyze and curate a robotics dataset using FiftyOne and Qwen3-VL embeddings, highlighting cross-embodiment patterns and methods for filtering near-duplicates.
Source: LeRobot — X · Read original article ↗
The robotics data ecosystem keeps getting stronger 🦾
@Voxel51 now reads LeRobot datasets natively, with episode visualization, embedding exploration and semantic search across embodiments 🤗
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
episodes
495
View original evidence
1. 495 of 497 clips have their nearest neighbor in the same session
Open source S5uniqueness score
292
uniqueness score
0.02 seconds
View original evidence
3. the two "most unique" clips are single-frame recordings, 0.02 seconds long
Open source S7episodes
497 episodes
View original evidence
i indexed all 497 episodes in fiftyone so you can search by text, cut the near-duplicates, and export a curated lerobot v3 subset
Open source S9Robot count
50 robots
View original evidence
i pulled 10 episodes from each of the 50 robot types in lerobot's community dataset and embedded every clip with qwen3-vl
Open source S4episodes
10 episodes
View original evidence
i pulled 10 episodes from each of the 50 robot types in lerobot's community dataset and embedded every clip with qwen3-vl
Open source S4
- dataset: not_reported
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: Available Artifact link Open source S12
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
i pulled 10 episodes from each of the 50 robot types in lerobot's community dataset and embedded every clip with qwen3-vl
Open source S4
1. 495 of 497 clips have their nearest neighbor in the same session
Open source S5
2. 292 score below 0.1 uniqueness
Open source S6
3. the two "most unique" clips are single-frame recordings, 0.02 seconds long
Open source S7
i indexed all 497 episodes in fiftyone so you can search by text, cut the near-duplicates, and export a curated lerobot v3 subset
Open source S9
read the dataset card: https://huggingface.co/datasets/Voxel51/community_v3_10per_embodiment
Open source S12
start here, read the full blog: https://huggingface.co/blog/harpreetsahota/fiftyone-now-reads-lerobot-50-embodiments-497-epis
Open source S11
you balanced your robot dataset by embodiment. the embeddings say you balanced it by recording session i pulled 10 episodes from each of the 50 robot types in lerobot's community dataset and embedded every clip with qwen3-vl 1. 495 of 497 clips have their nearest neighbor in the same session 2. 292 score below 0.1 uniqueness 3. the two "most unique" clips are single-frame recordings, 0.02 seconds long filter the session out and the cross-embodiment signal shows up: a cloth fold on one rig lands next to so100 arms folding cloth i indexed all 497 episodes in fiftyone so you can search by text, cut the near-duplicates, and export a curated lerobot v3 subset one prompt to an agent with fiftyone skills built the whole thing start here, read the full blog: https://huggingface.co/blog/harpreetsahota/fiftyone-now-reads-lerobot-50-embodiments-497-epis read the dataset card: https://huggingface.co/datasets/Voxel51/community_v3_10per_embodiment @LeRobotHF @Alibaba_QwenView the quoted post on X
Source：LeRobot · x.com