datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
messy_pick_object_place_white_bowlPick object from the frying pan and place it in the white paper bowl.
messy_pick_object_place_plat_spotPick [object from the green box/ egg from the large round plate] and place it in the frying pan.
ObjaversePlusPlus
Objaverse++: Curated 3D Object Dataset with Quality Annotations
Paper
Code
Chendi Lin,
Heshan Liu,
Qunshu Lin,
Zachary Bright,
Shitao Tang,
Yihui He,
Minghao Liu,
Ling Zhu,
Cindy Le
We cleaned the Objaverse dataset so you don't have to. In this work, we meticulously curated a collection of Objaverse objects and developed an effective classifier capable of scoring the entire Objaverse. Our extensive annotation system considers geometric structure and texture… See the full description on the dataset page: https://huggingface.co/datasets/cindyxl/ObjaversePlusPlus.ObjaNeRF-Textopenshape-objaverse-embeddingsobjaverse-lvis-1k-watertightrepro-causal-jepa-learning-world-models-through-object-level-latent-masking-traces
Agent traces
Agent sessions published from a Trackio Logbook.
object_directionBS-Objaverse
BS-Objaverse 660K Dataset Card
Dataset details
Dataset type:
BS-Objaverse 660k Dataset is a set of GPT4-Vision-powered multi-modal captions data.
It is constructed to enhance modality alignment and fine-grained visual concept perception for describing detailed information about the shape, texture of Objaverse 3D object.
obj_descript_gpt_10k.json is generated by GPT4-Vision.
objaverse660k_mvllava7b.json is generated by our MV-LLaVA trained on GPT4-Vision-generated data.… See the full description on the dataset page: https://huggingface.co/datasets/Zery/BS-Objaverse.uninavid-objnav-demoone_combo_object_directionindian-courtroom-objections-100k
Indian Courtroom Objections Dataset (100K)
A synthetic dataset of 100,000 Indian courtroom objection scenarios for training AI judge models. Each example contains a case premise, opposing lawyer's question, an objection with legal reasoning, and the judge's ruling (SUSTAINED or OVERRULED) with a one-line legal explanation.
Dataset Structure
Splits
train: 95,000 examples
test: 5,000 examples
Format (Conversational)
Each example follows the TRL… See the full description on the dataset page: https://huggingface.co/datasets/pkheria/indian-courtroom-objections-100k.hidden-objects
Hidden-Objects
Image-object pairs with localized bounding boxes for learning realistic object placement in background scenes.
Project page: https://hidden-objects.github.io/
Backgrounds: Places365
Schema
Field
Type
Description
entry_id
int64
Unique row identifier
bg_path
string
Relative path to background image in Places365
fg_class
string
Foreground object category (e.g. "bottle")
bbox
list
Normalized bounding box [x, y, w, h] in range 0–1
label… See the full description on the dataset page: https://huggingface.co/datasets/marco-schouten/hidden-objects.pi05-libero-plus-objects-layout-failures
pi0.5 LIBERO-plus Objects Layout Failures
Failure summary artifacts for evaluating TensorAuto/tPi0.5-libero on the LIBERO-plus Objects Layout perturbation subset through OpenTau.
Evaluation Setup
Benchmark: LIBERO-plus
Suite: libero_10
Perturbation category: Objects Layout
Policy: TensorAuto/tPi0.5-libero
Tasks: 312
Episodes per task: 5
Metric episodes: 1560
Seed schedule: episode seeds 1000 to 1004
Episode length: 520 steps
Source machine: wzxuan under… See the full description on the dataset page: https://huggingface.co/datasets/d3d3shan/pi05-libero-plus-objects-layout-failures.libero_plus_object
libero_plus_object: detailed LeRobot v3.0
This dataset was converted from the LIBERO Plus LeRobot v2.1 libero_plus_object partition.
The original 8D state and 7D action vectors are preserved exactly as
raw_state.ref_state and raw_action.ref_action. Canonical low-dimensional fields follow
failure_rollout_data/dataset.md; debug.gripper_eef_* contains the ground-truth next-step
relative EEF motion for inspection.
Required camera transform for canonical training
The… See the full description on the dataset page: https://huggingface.co/datasets/typoverflow/libero_plus_object.birds-object-detection-1600x896
Birds Object Detection 1600x896
A single-class bird object-detection dataset prepared for Ultralytics at a
1600 x 896 camera resolution. The dataset uses a deterministic 90% training
and 10% validation split.
Contents
data/yolo_birds_1600x896.tar.gz: complete Ultralytics dataset archive.
manifest.json: archive size, SHA-256 checksum, and dataset metadata.
After extraction, the archive contains data.yaml and matching image and YOLO
label trees:
data.yaml… See the full description on the dataset page: https://huggingface.co/datasets/PBatch23888/birds-object-detection-1600x896.E2E_real_object
E2E Real Object Direction
A video-based benchmark for evaluating VideoLLMs' directional reasoning and object recognition on real-world objects.
Conditions
Condition
Question
Answer
Purpose
direction_only
"In which direction is the object moving?"
"Up"
Baseline direction recognition
direction_obj_in_q
"In which direction is the car moving?"
"Up"
Does naming the object help?
direction_obj_in_a
"In which direction is the object moving?"
"The car is moving up"… See the full description on the dataset page: https://huggingface.co/datasets/KHUjongseo/E2E_real_object.object-distance-estimation-json
Object Distance Estimation Dataset
JSON dataset mapping detected objects to estimated distances
for robot vision systems.
objectnav-hm3d-metadata
objectnav-hm3d-metadata
Prepared metadata mirror for Habitat ObjectNav HM3D episode files.
Source:
Hugging Face dataset: Believe0029/HM3D
Source file: objectnav_hm3d_v2.zip
Contents:
val: 1000 episodes
val_mini: 30 episodes
This mirror contains episode metadata only. It does not include HM3D scene assets.
Use it to let em-eval auto-fetch ObjectNav split definitions while keeping scene
meshes and simulator assets local.
Swift_libero_objectobject_direction_1kSGP-Objectobjaverse-xl-shape-annotations
objaverse-xl-shape-annotations
Shape-based textual annotations for 537,841 objects from Objaverse-XL. Each object gets a class label and a short, geometry-focused description.
Why does this exist?
Objaverse-XL is a large benchmark, but it does not contain any textual descriptions.
This dataset was built to fix that. Every description focuses strictly on shape and structure, making it suitable for text-to-3D retrieval and contrastive representation learning tasks… See the full description on the dataset page: https://huggingface.co/datasets/valira-ai/objaverse-xl-shape-annotations.wikidata_rdf_massive_objects_ENobject_location_4class_video
Object Location 4-Class (Subset 1 of 5)
A visual question answering dataset for evaluating spatial reasoning — specifically, identifying the location of an object (Up / Down / Left / Right) from the viewer's perspective.
Dataset Configurations
close_ended
Multiple-choice format with 4 fixed options (A: Up, B: Down, C: Left, D: Right).
open_ended
Free-form answer format. The model is expected to respond with one of: Up, Down, Left, Right.… See the full description on the dataset page: https://huggingface.co/datasets/KHUjongseo/object_location_4class_video.libero_object_mj332
libero_object_no_noops_lerobot: detailed LeRobot v3.0
This dataset was converted from the Fast-WAM LIBERO MuJoCo 3.3.2 LeRobot v2.1
libero_object_no_noops_lerobot partition. It contains successful demonstrations whose historical no-op actions
were removed by simulator replay before this conversion. This converter preserves all remaining
frames and does not apply any additional filtering.
The original 8D state and 7D action vectors are preserved exactly as
raw_state.ref_state and… See the full description on the dataset page: https://huggingface.co/datasets/typoverflow/libero_object_mj332.bbh-logical-deduction-seven-objects-plobject_location_4class
Object Location 4-Class (Subset 1 of 5)
A visual question answering dataset for evaluating spatial reasoning — specifically, identifying the location of an object (Up / Down / Left / Right) from the viewer's perspective.
Dataset Configurations
close_ended
Multiple-choice format with 4 fixed options (A: Up, B: Down, C: Left, D: Right).
open_ended
Free-form answer format. The model is expected to respond with one of: Up, Down, Left, Right.… See the full description on the dataset page: https://huggingface.co/datasets/KHUjongseo/object_location_4class.han-humanoid-vision-object-detection-metrics-v1
Humanoid Vision Detection Metrics Dataset
Overview
Dataset performa sistem visi humanoid saat mendeteksi objek.
Features
image_brightness_level
object_distance_cm
detection_confidence_score
camera_noise_index
frame_processing_time_ms
Target
detection_accuracy_percent
cybersec-jsonschemabench-cloudtrail-objective-natural-hard-v5
CybersecJSONSchemaBench CloudTrail Objective Natural Hard v5
This is a 100-problem natural-prompt long-context cybersecurity reasoning subset built from the full flAWS CloudTrail corpus.
Each row contains a natural analyst-style question, a large CloudTrail JSONL context, and the JSON schema the answer must match. Gold answers are deterministic hidden-oracle results over the serialized slice and are not included in this public export.
Families
actor_recon_to_change: 22… See the full description on the dataset page: https://huggingface.co/datasets/achinta3/cybersec-jsonschemabench-cloudtrail-objective-natural-hard-v5.