Ego4D: First-Person Video with Feature Extraction Tools

Ego4D is an egocentric video repository developed by Facebook Research, featuring tools for feature extraction and visualization as of March 2026. It provides a framework for processing first-person video data for computer vision applications, though specific record counts are not listed in the repository metadata. The project includes example usage scripts to assist researchers in implementing egocentric video models.

Use Cases

• Extracting visual features from video data using the provided extraction scripts
• Visualizing egocentric video sequences using the built-in visualization tools
• Implementing computer vision workflows using the provided example usage scripts

Strengths

• Includes dedicated feature extraction and visualization tools
• Provides example usage scripts for implementation
• Released under the MIT license

Limitations

• Repository metadata does not specify the total number of records or dataset size
• Requires external download scripts to access the primary video data files

Provenance

Source

facebookresearch

Freshness

Updated as of March 2026.