OpenCV is the standard for computer vision. However, it was not made to be put on an Nvidia Jetson board. Nvidia is trying to fix this, and have created an easy-to-use SDK installer. However, this installer is in beta. It was very easy to pick and choose which libraries to install on my board until the SDK would not properly connect to the board. Unfortunately, I had to revert to a much more complicated way which I figured out by compiling multiple advice pages on the internet. In the end, I got OpenCV working by getting a simple python script to display the video feed coming from a USB camera plugged into the Jetson.
Originally, I planned on continuing to work on the app. However, I realized that my work with the GPIO pins that I had previously done was useless because it was written in C++. Previously, I used a GPIO library I found for the Jetson TX1 which I edited for the TX2. However, I was not able to find a GPIO library in Python. Therefore, I had to write the GPIO library from scratch including functions such as Export, Unexport, setDirection, SetValue, GetValue, and ActiveLow.
My progress is slightly behind because of having to write the GPIO library in Python.
My goal for this week is to get Jing’s graph integrated with TensorRT. After some preliminary research, this does not seem straightforward as I had hoped.