MAVOS-DD multilingual deepfake detection benchmark
An open multilingual audio-video dataset for deepfake detection.
- Video classification
- Arabic
- Romanian
- English
- +4
Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.
Answered by Gemini 4 · the chat is kept only in this tab
An open multilingual audio-video dataset for deepfake detection.
A large corpus of human action video, audio, and 3D data for robotics and embodied AI.
An English video corpus for training models to answer questions about videos.
Media data and code agent run logs for the Embodied First Exam robot simulation benchmark.
96 single-camera driving videos and over 105 thousand frames with depth and semantic segmentation annotations.
10,000 hours of real human work video with head and hand pose annotations in video format for robotics.
A filtered selection of 10,000 to 100,000 dynamic video clips without static scenes for training world models and video generation.
English video QA dataset for evaluating multimodal neural network video understanding in real time.
About 400k English video clips with detailed scene, shot, and camera motion descriptions.
100M to 1B entries featuring video, time series, and robot joint states.
A video dataset generated by an agent framework for pre-training Vision-Language-Action models.
Over 500 hours of R1-Lite mobile robot video with action annotations for robotics training.
Multimodal dataset for robotics: 1,693 episodes, 273 thousand video frames, and telemetry in parquet.
100 hours of egocentric camera and tactile glove recordings for robotic manipulation training.
Over 40k Q&A pairs for evaluating multimodal models in video, image, and text anomaly detection.
Archive of real robot execution runs featuring videos, success scores, and preferences for RLHF and VLA models.
Videos and simulations of robot work in chemical laboratories for training and evaluating manipulation skills.
Over 215 hours of video from 41 cameras and tactile interaction data with 198 deformable objects.
Over 100k first-person CS2 rounds with 720p video, stereo audio, and tick-synchronized data for all 10 players.
4,320 audio-visual videos from 15 accelerated versions of the MiniMax-H3 model for evaluating generation quality in few steps.
A selection of 28,562 video clips filtered via an allowlist from the large HOIGen-1M dataset.
A large dataset (100K-1M records) with video, sensor readings, and clothing folding annotations in real homes.
50 robot pick-and-place episodes for the LeRobot framework, including video and telemetry.
Over 108,000 short video clips across 2,208 American Sign Language words for sign recognition.