4_28_22: Modeling Updates Meeting

Updates:

• Sumaiya completed the training for logistic regression and random forest is in progress. Working on evaluation functions to be used to gauge performance of models.
• Shubham completed deep learning configuration. Currently using demo code to fine tune training the neural network, and is testing images resulted from video stream.
• Allen completed scripts to take raw images from video stream and feed into pipeline to prepare for model predictions. Working on coding additional scripts.

Issues/Next Steps:

• XGBoost code needs to be troubleshooted, since the randomized grid search does not result as expected.
• Need to adjust images extracted from video feed to be more similar to the training images. Initial testing shows that model performance on the extracted images is not meeting our expectations. It is determined that the extracted images include more background than training images.
• Finalize modeling and evaluation. Continue coding pipelines and scripts.