I am currently working on deploying a simple object detection model trained on yolov7. I exported the model as a .onnx type and then converted it to .blob. While running the pipeline on our OAK-D Pro Wide the detections fed from the neural network queue are not lining up with the expected value. I am getting confidences up to 3.3, labels of 0 and 1 when there is only one label, and negative x and y values.
Example Output:
Label: 0
Confidence: 1.1123046875
Xmin: -8.9375
Ymin: 0.10968017578125
Xmax: -1.23828125
Ymax: 1.1015625
