FishSense Post-Correction Utility

This utility provides the tools to apply affine corrections to a RealSense bag file

Overall process

1. Identify temporal areas of interest.
2. Extract the component streams from the ROSBAG (RGB video, depth video, sensor transforms)
   1. Execute C:\Program Files (x86)\Intel RealSense SDK 2.0\tools\rs-convert.exe -c -d -i ${rosbag} -p ${png_prefix}
3. Apply color correction
4. Label the RGB video
   1. Create a new Label Studio Project
   2. Add the data
   3. Set the labeling template as label-studio_template.xml
   4. Set the labeling instructions to the following:

   Mark the tip and tail of each fish first.  Then, apply a rectangle over each fish.
   
   Only select fish which are completely flat and clearly visible.  This is not an exercise in finding every fish, only selecting the measurements that we desire to make.
   
   You must select the tip and tail of each first first before creating the rectangle.  The rectangle must encompass exactly the tip and tail of the fish - we don't care to have a tight box around the fish.
   

5. Match the RGB video with depth data
6. Extract corrected labels

Dependencies

Name	Version	URL
Intel RealSense SDK	v2.51.1	https://github.com/IntelRealSense/librealsense/releases/tag/v2.51.1
Python 3	3.9.13	https://www.python.org/downloads/release/python-3913/

File formats

Temporal Areas of Interest

This file will be a text file denoting temporal areas of interest, one per line.

Temporal areas of interest shall be defined by a start and end time, referenced to the start of the clip.

For instance, if we want to work on the clip 160.bag, the file shall be named 160.bag.times.txt

If we want to denote the ranges 13:11.012-15:11.016, 17:12.561-19:32.781, we would put the following in the file:

00:13:11.012-00:15:11.016
00:17:12.561-00:19:32.781

Formally, the timestamps shall be in hh:mm:ss.sss. If possible, comply with ISO 8601 timestamp formats and include as much precision as possible.

Color Correction

Developer Setup

1. Open fishsense-postcorrection.code-workspace in VS Code
2. In the root of this repository, run python -m venv .venv to create the virtual environment
3. Activate the virtual environment
   1. On Windows, run .venv/Scripts/Activate
4. Install the repository as editable by running python -m pip install -e .