This week we were finally able to get our hands on a microSD card so we can flash the jetson and get things started when it comes to running code on the board. Our plan is to get the main architecture running and then load a set of random weights for the CNN and see how much memory is used and if we are going to be able to run our code on multiple threads. We are also going to have to set up our code so that the main script is run on the cpu while the CNN calculations are to be run using the GPU. On the topic of the CNN, I was able to get the training code ready to go for the model and all I have to do now is set up an ec2 instance along with a neptune dashboard to train and track the performance of the training. I said in my personal report that I am hoping to have the model ready to be trained by Friday this week.
Also because the SD card finally came in we were able to meet in person and flash the jetson board and starting processing the video stream with openCV. We were able to flash the board and get it up and running, since we were waiting for a SD card larger than 32gb. We were also able to setup the camera as a test, to ensure the camera was working. Now that these components can be connected correctly we then can move forward from this point and start deploying all of the models we need. We were able to start on PyTorch however, it there was an issue that needed the Nano to be adjusted. But we were able to look into it with some research and fix, the final deployment will be done this upcoming week for the untrained models.
0 Comments