anno2trackstart

Originally designed for siphonophore behavior research, this repo contains code to turn MBARI VARS annotations into video links to the start of an animal sighting, using YOLOv5-assisted tracking

Install

1. Clone respository with submodules

git clone --recurse-submodules https://github.com/bioinspirlab/anno2trackstart.git

2. Set up python environment using Python 3.8
3. Open the 'anno2trackstart.ipynb' file and set the kernel to the python environment
4. Ensure that pip and ipykernel are installed.
5. Run the code to 'install dependencies' and 'import libraries'

Note: If recieving errors around Lap installation, downgrade pip to version 20.0.2

Run

1. Add the path to the csv file containing the annotation dataset and run

4. Run part_one to find, trim, and track the annotations and videos

The output of part one will be a folder in the working directory containing trimmed, untracked videos, and tracked videos with their associated tracking txt files

3. Set the variables for part two with the path to the tracked files, and the terms used to find t0

Example:

4. Run part two to find t0 and create easy access link based on the t0 timestamp and original videos