DF26 talking-head deepfake benchmark
2,691 short HD videos (real and generated) for evaluating deepfake detectors.
- Video classification
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
2,691 short HD videos (real and generated) for evaluating deepfake detectors.
A benchmark of English videos for detecting and localizing precise moments of audio and video tampering.
Large English video dataset with temporal action segments and text descriptions.
From 10k to 100k records with video, joint poses, and finger poses of the Unitree G1 robot during cushion pickup.
Video data for training and evaluating the Goku text-to-video generative model.
Parameters and 3D head meshes including teeth and tongue for 15,345 multi-view frames of 256 individuals.
1,313 examples with 74.8 thousand human ratings for videos generated by the Marey model.
102 hours of first-person viewpoint video for training robotics and embodied AI models.
Annotated English video calls from NVIDIA for evaluating conversational multimodal agents.
6-axis manipulator motion recordings in Isaac Lab simulator for robot object grasping training.
262 dual-arm robot manipulation episodes (30 fps video and telemetry) in LeRobot and parquet formats.
Raw sensor data and onboard video from five overhead transport systems of the KAIST laboratory, unannotated.
Expert benchmark with programmatic verification to evaluate how LLMs animate vector graphics.
Teleoperated robot episode samples for warehouse tasks in English.
Recordings from two stereo endoscopes (~48 GB) for depth estimation, pose, and 3D reconstruction in surgery.
96 high-resolution, silent video recordings from a fixed camera at a honeybee hive entrance.
Dataset for evaluating humanoid robot safety in logistics warehouses with incident annotations.
10,000 hours of synchronized head and wrist video with 3D hand poses for robot learning.
69 six-second clips from the classic animated short with prompts for video generation.
A set of 1,000 short videos with detailed manual annotation of events, actions, and movements.
Two hours of gameplay video from 24 games with millisecond-accurate keypress, mouse movement, and layout recordings.
The first human-oriented benchmark for evaluating physical realism in videos.
Set of 87.7k short video clips with three descriptions each for evaluating multimodal retrieval in long videos.
Egocentric videos with synchronized 3D hand poses for training and evaluating video models.