I did research about algorithms to consider for the Detection phase which involves pointing out the location of the subject with bounding boxes around it. Refreshed my knowledge on the way CNNs work and then weighted out the pros and cons of different algorithms to narrow down 3 that we will be testing and evaluating on training datasets. The attached file shows the summary of the research and includes the URL’s of websites that would be a reference for when the algorithms are being implemented.
Apart from this, I was able to select and download a large dataset of images of animals that would be our training data for the models. I began setting up the environment, etc. for the YOLOv5 algorithm and worked on formatting the images for being inputted here.
My progress on the project is a little behind however it should not be difficult to make up time. To catch up I will begin testing and evaluating the models as soon as possible and do this while simultaneously working on building the physical setup during lab time (as planned for by the schedule)
I plan to pick an algorithm as well as have data to support the reason for doing so. Along with this we expect to have most of the physical setup ready and be able to relay the real-time from the camera onto the computer, using the Jetson.