The project that I made for my dissertation titled. Action Detection in Untrimmed Videos.
The action_model.py file is a suite with functions that can be used for manipulating the VIRAT dataset
This project uses YOLOv7 "https://github.com/WongKinYiu/yolov7" for the object detection and SORT for object tracking "https://github.com/abewley/sort".
Where in the code it says, pathtoYOLOv7. The YOLOv7 github will need to be cloned, see the address above, and the path to the 'yolov7-main' file location will need to be replace 'pathtoYOLOv7'
One video has been provided from the VIRAT dataset (as providing all the dataset is too large), it is under '\VIRAT Ground Dataset\videos_original\VIRAT_S_000006.mp4''
The full dataset can be downloaded from: https://viratdata.org/