I was originally hoping that Jing’s ML algorithm and Irene’s CV algorithm would smoothly work on the Jetson. I had prepared for this by installing all the necessary libraries for Tensforflow and OpenCV, in addition to the TensorRT libraries. However, after attempting to run the Python script on the Jetson there was a library mismatch. The drivers for the GPUs were too old of a version to work with the Python script. Unfortunately, there is no easy way to update the drivers. Originally, I was using the OS that boot up out of the box. Now, I had to download the most recent OS and re-flash the whole board.
After the re-flashing of the board, I had to reinstall all of the libraries. This process was a lot quicker than previously because I already knew what I had to do, but was rather tedious. Finally, Jing and I started to try and run the Python script. The Python script ran perfectly well on his PC laptop, now we just had to get it to run on the Jetson. After updating to the newest version of Tensorflow, and modifying bits and pieces of the CV and ML portions of the code, we were finally able to get it to work! It ran very fast, although this conclusion is just from my eye’s view, we have yet to run tests.
My progress is behind due to the complications that arose in running the Python script on the Jetson. To catch up, I will not be focusing my time on optimizing the algorithm for the Jetson, as this is not a necessary aspect of the project, simply a desired aspect.
In the next week I will be working on fully integrating the entire system. After this is accomplished I hope to be running a variety of tests on the system.