We used the D435 Intel RealSense Depth sensor to recognize balls. We performed several steps in order for the robot to recognize a ball.

Steps

1. Created a ball detection model: a set of all the weights and biases that make up the network
2. Exported the model as a .onnx file
3. Loaded the model onto the Jetson
4. Used libraries that were already on the Jetson to use the model to make predictions on images
5. Every time the robot runs, the depth sensor is constantly capturing images and making predictions with the model.