Ego4D: 3,025 hours of egocentric video captured in 74 cities in 9 countries by 855 participants - a consortium of 13 universities in partnership with Facebook AI Research https://ego4d-data.org/ **Data available end of Nov 2021**
EPIC-KITCHENS: 20M Frames, full HD, head-mounted, 45 kitchens from 4 cities (in North America and Europe). FULLY annotate.
Dataset from: http://epic-kitchens.github.io
The dataset is the first of its kind, in size, annotations, diversity and unscripted nature of its collection. It is accompanied by challenge leaderboards to allow comparative evaluation of methods on 6 benchmarks. Since its first-version release in 2018, the dataset has: 2.4K unique downloads from 54 countries (most downloaded dataset from UoB data.bris), 1.5K submissions to challenge leaderboards (by May 2021), 410 peer-reviewed citations.
C++ implementation for capturing and registrating 3D models from two frontly facing Kinects [Author: Vahid Soleimani].
C++ Linux-based code for real-time learning and detection of texture-minimal objects.
MATLAB code is available with details of how to use it download (3.0MB)
A short sequence from PETS2006 to test the code - download (3.9MB)
C++ code coming soon...