Daily Tasks to-do

08/02/2024

Tasks to do in parallel:

• Talk to Jake about getting tasks up on JIRA.
• Finish running single video through CCTV CV pipeline.
• WandB integration into CCTV ML.
• Get WSL set up and loe-coe app running on my msi laptop.
• Get CUDA working with light gbm.
• Creating blueprint for new utility for CCTV CV.
• Training the model with both SD1_B data and 2023_05_26 data.

08/13/2024

Discussion with Sudhir

1. For the fastai_multi_label_predict.py script, uncomment the lines for using the test folder. Since the script is based on the folder where images are stored there are two approaches that can be taken:

• Copy test_images to another folder.
• Look into using get_x() to get images specifically for the test frames from the test_images folder.

2. ~~Create JIRA task to document Streamlit app in WSL issue with the NVIDIA driver. ~~
3. Unpivot the predictions table so that there are three columns: frame_id, defect, probability_of_defect. Do this in the dataframe itself.
4. Create JIRA task to beef up the shell script so that you don't need to use dummy arguments.
5. Create two seperate docs files for the trained models for SD1.
6. Look into calling shell script from the streamlit app.
7. Change data_type in the train csv to use_type. Change fname to frame_name
8. Improve shell script to accept optional arguments.

08/14/2024

T7 remaining space: 416 GB Blurred Frames folder with zips: 54GB. Unzipped might be a bit larger.

08/15/2024

Started running extract SD1_B.