📊 Datasets — Awesome Robotics

This dataset was created using LeRobot. Dataset Structure meta/info.json: { "codebase_version": "v3.0", "robot_type": "panda", "total_episodes": 1693, "total_frames": 273465, "total_tasks": 40, "chunks_size": 1000, "data_files_size_in_mb": 100, "video_files_size_in_mb": 500, "fps": 10.0, "splits": { "train": "0:1693" }, "data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet", "video_path":… See the full description on the dataset page: https://huggingface.co/datasets/lerobot/libero.

A benchmark of 50 robotic manipulation tasks for multi-task and meta-reinforcement learning.

RLBench is a dataset and benchmark that contains a variety of continuous manipulation tasks used to evaluate robotic control and planning in cloud-robotic systems.

MuJoCo is a simulation environment that contains models for evaluating the performance of robotic systems, specifically used in this paper to demonstrate the effectiveness of a tendon-based control method for a high-dof anthropomorphic hand model.

'RoboTwin 2.0' is a benchmark dataset used to evaluate robotic manipulation tasks, specifically focusing on spatial-temporal interactions and geometric constraints in action execution.

Dataset Card for "clavin" More Information needed

The 'RoboTwin' dataset/benchmark contains simulation environments for bimanual manipulation tasks and is used to evaluate data-driven learning approaches in robotics.

'SimplerEnv' is a simulation benchmark used to evaluate continuous-action vision-language-action models in various settings, including standard, few-shot, and noisy conditions.

The 'Unitree G-1' dataset/benchmark contains motion data for humanoid robots and is used to evaluate the effectiveness of cross-embodiment transfer in whole-body tracking models.

HM-3D is a dataset that contains a collection of 3D scenes used to evaluate floor plan-guided embodied navigation tasks, including PointNav, ObjectNav, and ImageNav.

LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Models 📄 Paper | 🏗️ Repo | 🌐 Website | 🤗 Assets | 🤗 Model | 📁 Training Dataset 🔥 Overview This repository contains the official implementation and benchmark for our paper "In-depth Robustness Analysis for Vision-Language-Action Models". We systematically expose the hidden vulnerabilities of contemporary VLA models through comprehensive robustness evaluation across seven perturbation… See the full description on the dataset page: https://huggingface.co/datasets/Sylvest/LIBERO-plus.

RoboCasa Dataset for ManiSkill/SAPIEN This is the robocasa dataset from RoboCasa, it has been adapted for use with ManiSkill/SAPIEN. The changes are: Re-exported some material.png files to fix colors. Some materials showed up incorrectly as red. Add some missing default textures for cabinets/panels re-exported materials list: wall white_bricks material outlet materials marble_5 all toaster materials/images fridgeaire_gas materials 1_bin_storage_right_dark materials stool_1_3

The 'RoboMimic' dataset/benchmark contains a variety of robotic manipulation tasks and is used to evaluate the performance of robotic control algorithms in real-time scenarios.

ManiSkill Data ManiSkill is a unified benchmark for learning generalizable robotic manipulation skills powered by SAPIEN. It features 20 out-of-box task families with 2000+ diverse object models and 4M+ demonstration frames. Moreover, it empowers fast visual input learning algorithms so that a CNN-based policy can collect samples at about 2000 FPS with 1 GPU and 16 processes on a workstation. The benchmark can be used to study a wide range of algorithms: 2D & 3D vision-based… See the full description on the dataset page: https://huggingface.co/datasets/haosulab/ManiSkill.

The 'R-2R' dataset/benchmark contains navigation tasks that require agents to follow natural language instructions in photo-realistic environments and is used to evaluate the robustness and effectiveness of vision-and-language navigation methods.

The MP-3D dataset is used to evaluate Open-Vocabulary Object Navigation (OVON) by providing environments for embodied agents to locate language-specified targets.

The 'Adroit' dataset is a benchmark used to evaluate robotic manipulation tasks, specifically focusing on visuomotor control.

The 'Gazebo' dataset/benchmark is a simulation environment used to evaluate the effectiveness and versatility of control frameworks for legged robots, specifically in terms of stability and robustness against uncertainties and disturbances.

IsaacSim is a simulation platform that contains high-fidelity models for evaluating the performance of aerial robots in navigation tasks, specifically in the context of object-goal navigation.

The 'LIBERO-LONG' dataset/benchmark contains data for evaluating robotic manipulation tasks, specifically focusing on the effectiveness of planning frameworks like CLaD in achieving successful kinematic and semantic transitions.

The Replica dataset is a benchmark that contains a collection of indoor scenes used to evaluate the performance of SLAM systems in terms of 3D geometry and camera tracking accuracy.

Dataset Card for REVERIE Dataset Details Dataset Type: REVERIE is the first large-scale visual instruction-tuning dataset with ReflEctiVE RatIonalE annotations. REVERIE comprises 115k machine-generated reasoning instructions, each meticulously annotated with a corresponding pair of correct and confusing responses, alongside comprehensive rationales elucidating the justification behind the correctness or erroneousness of each response. Data Collection: REVERIE was… See the full description on the dataset page: https://huggingface.co/datasets/zjr2000/REVERIE.

The 'VLN-CE' dataset/benchmark contains language instructions paired with visual observations for evaluating embodied navigation tasks that require translating these inputs into spatial actions for robots.

ALFWorld is a benchmark dataset that contains multimodal environments used to evaluate the performance of embodied agents in perception and decision-making tasks.

The 'DROID' dataset/benchmark is used to evaluate visual prediction models for embodied control by providing a collection of robotic manipulation tasks and corresponding visual states.

Isaac Gym is a benchmark that contains a diverse set of simulated robotic environments and tasks used to evaluate the performance of Vision-Language-Action models in cross-embodied settings.

'SIMPLER' is a benchmark dataset used to evaluate Vision-Language-Action frameworks in the context of robot manipulation tasks.

The 'Unitree Go-2' is a quadruped robot used to evaluate safe locomotion in unstructured environments, focusing on long-horizon goal progress, passability over uneven terrain, and collision avoidance against high-speed dynamic obstacles.

The Matterport-3D dataset is a collection of 3D indoor environments used to evaluate autonomous navigation algorithms.

The 'R-2R-CE' dataset/benchmark is used to evaluate vision-and-language navigation systems by providing a set of tasks that require grounding natural-language instructions into navigation actions in complex environments.

This dataset was created using LeRobot. Dataset Structure meta/info.json: { "codebase_version": "v2.0", "robot_type": "unknown", "total_episodes": 206, "total_frames": 25650, "total_tasks":1, "total_videos": 206, "total_chunks": 1, "chunks_size": 1000, "fps": 10, "splits": { "train": "0:206" }, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path":… See the full description on the dataset page: https://huggingface.co/datasets/lerobot/pusht.

This is ScanNet dataset containing indoor scenes which is used for 3d object detection, 3d segmentation, etc. Acknowledgement @inproceedings{dai2017scannet, title={ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes}, author={Dai, Angela and Chang, Angel X. and Savva, Manolis and Halber, Maciej and Funkhouser, Thomas and Nie{\ss}ner, Matthias}, booktitle = {Proc. Computer Vision and Pattern Recognition (CVPR), IEEE}, year = {2017} }

ALFRED is a benchmark dataset that contains tasks for evaluating robotic planning and decision-making capabilities in real-world scenarios.

COLOSSEUM is a benchmark used to evaluate the performance of robotic manipulation tasks, specifically assessing the effectiveness of 3D-aware visual pretraining methods.

The GOAT Benchmark (HomePage) We introduce the GOAT-Bench, a comprehensive and specialized dataset designed to evaluate large multimodal models through meme-based multimodal social abuse. GOAT-Bench comprises over 6K diverse memes, encompassing a range of themes including hate speech and offensive content. Our focus is to assess the ability of LMMs to accurately identify online abuse, specifically in terms of hatefulness, misogyny, offensiveness, sarcasm, and harmfulness. We… See the full description on the dataset page: https://huggingface.co/datasets/HKBU-NLP/GOAT-Bench.

GraspNet-1Billion (Unofficial Mirror) ⚠️ DISCLAIMER / 声明: This is an UNOFFICIAL mirror/upload of the GraspNet-1Billion dataset. This repository is maintained solely for ease of access and research reproducibility. I do not own the rights to this data. All rights belong to the original GraspNet team, MVIG, and SJTU. 🔗 Official Sources / 官方源 Please always refer to the official website for the latest updates, and documentation: Official Website:… See the full description on the dataset page: https://huggingface.co/datasets/DravenALG/GraspNet-1Billion.

Dataset Card for Kitti The Kitti dataset. The Kitti object detection and object orientation estimation benchmark consists of 7481 training images and 7518 test images, comprising a total of 80.256 labeled objects

'ManiSkill-3' is a benchmark dataset that evaluates robotic manipulation tasks using egocentric visual perception, focusing on the challenges of occlusion, motion blur, and partial observability in dynamic environments.

The 'RxR-CE' dataset/benchmark is used to evaluate Vision-and-Language Navigation (VLN) systems by providing a collection of navigation tasks that involve understanding and interpreting visual and linguistic information in complex environments.

VLABench is a benchmark dataset used to evaluate the performance of Vision-Language-Action models in detecting execution failures during robotic task execution.

'BEHAVIOR-1K' is a benchmark that evaluates the performance of Vision-Language-Action models in embodied tasks.

Dataset Card for "calvin_abc_d" More Information needed

Dataset of "DexGraspNet 2.0: Learning Generative Dexterous Grasping in Large-scale Synthetic Cluttered Scenes" (CoRL 2024) Code: https://github.com/PKU-EPIC/DexGraspNet2

The 'Gibson' dataset is a benchmark used to evaluate robotic navigation methods, specifically focusing on the ability to navigate through complex environments.

The 'HM-3D-OVON' dataset/benchmark contains 3D environments and is used to evaluate the performance of Vision-Language Navigation agents in their ability to follow instructions and navigate efficiently.

The 'Isaac Lab' is a simulation environment that contains parallel environments for training reinforcement learning policies, specifically designed to evaluate energy-efficient actuation strategies for hybrid aerial-ground robots navigating stair-like terrain.

MimicGen is a dataset/benchmark that contains a variety of manipulation tasks used to evaluate the performance of robotic policies in simulated environments.

Open X-Embodiment Dataset (unofficial) This is an unofficial Dataset Repo. This Repo is set up to make Open X-Embodiment Dataset (55 in 1) more accessible for people who love huggingface🤗. Open X-Embodiment Dataset is the largest open-source real robot dataset to date. It contains 1M+ real robot trajectories spanning 22 robot embodiments, from single robot arms to bi-manual robots and quadrupeds. More information is located on RT-X website… See the full description on the dataset page: https://huggingface.co/datasets/jxu124/OpenX-Embodiment.

The 'PyBullet' dataset/benchmark is a simulation environment used to evaluate robotic systems, specifically assessing their performance in various tasks such as re-planning, failure recovery, and policy enforcement.

The 'SimplerEnv-WidowX' is a simulation benchmark used to evaluate robotic manipulation performance in the context of Vision-Language-Action frameworks.

The 'SOON' dataset is a large-scale benchmark derived from web-based room tour videos, used to evaluate Vision-and-Language Navigation by providing diverse, realistic indoor settings with enriched spatial and semantic supervision.

The YCB dataset is a benchmark that contains a collection of objects used to evaluate robotic manipulation and pose estimation methods, particularly in scenarios where visual perception is limited.

The D-4RL dataset/benchmark contains various reinforcement learning tasks designed to evaluate the performance of algorithms in offline reinforcement learning settings, specifically in environments like maze2d and antmaze.

This repository contains the EmbodiedOcc-ScanNet dataset, which is a reorganized benchmark based on local annotations, designed to facilitate the evaluation of the embodied 3D occupancy prediction task. It accompanies the paper EmbodiedOcc: Embodied 3D Occupancy Prediction for Vision-based Online Scene Understanding. Project page: https://ykiwu.github.io/EmbodiedOcc/ Code: https://github.com/YkiWu/EmbodiedOcc Paper Abstract 3D occupancy prediction provides a comprehensive… See the full description on the dataset page: https://huggingface.co/datasets/YkiWu/EmbodiedOcc-ScanNet.

The 'Habitat' dataset/benchmark contains realistic environments and scenarios used to evaluate mobile manipulation tasks in robotics.

OgBench: Benchmarking Graph Neural Networks on Omics Data OgBench is the first benchmark suite for graph-level prediction in the n ≪ p regime characteristic of omics data, where the number of patient samples n is much smaller than the number of nodes (genes or proteins) p per graph. Datasets This repository contains four preprocessed omics graph classification datasets: Dataset Modality n p Task HERITAGE Proteomics 654 4,977 Exercise responder… See the full description on the dataset page: https://huggingface.co/datasets/geometric-intelligence/ogbench.

The 'VGGT' dataset/benchmark is used to evaluate the visual-kinematic alignment of generated robot trajectories in diverse manipulation environments.

VirtualHome is a benchmark dataset used to evaluate Large Language Models (LLMs) on tasks related to understanding goals, planning actions, and executing tasks in simulated environments.