This week we were scrambling a little it because of how quick the turnaround was from proposals to an actual design. About halfway through the week we met as a team and got some ideas flowing for our block diagram and some of the smaller implementation details. We were essentially breaking our larger block diagram down into more detail for our written design review. We also spent some time talking about our schedule and how things were moving and if any tasks needed to be broken up or pushed back. One thing we decided needed to be broken up was actually creating our training data, we decided to break this up because the process for creating this data is looking like it may take more time than we originally thought.
One a separate note, I started playing around with some code in python using pytorch and tensorflow. I was not really familiar with pytorch but I have heard good things and wanted to make sure our decision to use tensorflow was justified, and after playing around with both I think our decision was justified. Tensorflow was easier to use and its syntax made more sense, which is important when most of your team is fairly new to ML. Unfortunately I did not have a large enough dataset to notice any performance differences but I have read that on average tensorflow is slightly faster so I think we made the right decision. Once we were set on tensorflow I started to get coding on the classification CNN because this is just a straightforward AlexNet encoder for image classification. It is time to start getting down to implementation time and I know I am excited to finally get coding!
0 Comments