I'm currently working on an project where the goal is real-time object detection. I have already trained a yolov8n model on my pc and it works well on my SBC too, to get inference that is. For consistent and repeatable testing, I am using pre-recorded sensor data from a ROS bag as the input, which streams images directly to the OAK device. I have successfully converted my yolo model from PyTorch (.pt) to the .blob format required by the OAK hardware's VPU.
I'm trying to implement this using the oak compute, but I am getting multiple errors. I would greatly appreciate any advice. I just need to be able to convert my pre-trained custom model into one that can be used on this board. Is there a more robust, known-working method for deploying custom YOLOv8 models on OAK cameras that I might be overlooking.
