Nathan’s Status Report 04/03

This week we decided that it was not worth it to wait for the original SD card that we ordered, so we put a request in for another SD card of smaller memory from Amazon, a much more reliable source than Wish.com, and the SD card came the next day. So once the SD card came in we were able to pass off the elements of the system and get the nano flashed and start testing some code on it. Rebecca has been handling most of the hardware, so while she was flashing the jetson I was still working on training the model. I have most, if not all of the training code ready to go so now I have to split the images up into 128×128 patches and then start training. My plan for training the model is to first try and run one training set of ~100 images and then 1 testing set and 1 validation set. I want to then log some accuracy metrics so I can track the progress and performance of training the model. Once I have this working on a small scale I am going to train the model using an ec2 container that will also be linked up to aws Neptune’s dashboard. This way I can offload the training onto an ec2 container so it doesn’t take up all of my laptop’s performance, and also monitor the performance of the model using Neptune. This model will probably take close to 3 full days to train and test, so my goal is to have the neptune dashboard and ec2 instance running by the end of the week.