Hi guys. Iam using OAK-D POE. I tried to to run the example code provided in SpectacularAi (depthai_combination.py) that uses depthai library. I tried to use my custom model (file.blob), by running the code. After running the code with new model, I got issue * Input image (416x416) does not match NN (640x640) * Mask is not defined for output layer with width '6'. Define at pipeline build time using: 'setAnchorMasks' for 'side6'. * [critical] IMU driver failed with error code 1. Your board likely doesn't have IMU! Does anyone know how to solve this ? and any reference relate to setNumClasses, setCoordinateSize, setAnchors, setAnchorMasks, dan setIouThreshold provided in https://docs.luxonis.com/projects/api/en/latest/components/nodes/yolo_detection_network/

Error when trying to run the new model instead of tiny yolov4 model

jenanaputra

Hi guys. Iam using OAK-D POE. I tried to to run the example code provided in SpectacularAi (depthai_combination.py) that uses depthai library. I tried to use my custom model (file.blob), by running the code. After running the code with new model, I got issue

Input image (416x416) does not match NN (640x640)
Mask is not defined for output layer with width '6'. Define at pipeline build time using: 'setAnchorMasks' for 'side6'.
[critical] IMU driver failed with error code 1. Your board likely doesn't have IMU!

Does anyone know how to solve this ? and any reference relate to setNumClasses, setCoordinateSize, setAnchors, setAnchorMasks, dan setIouThreshold provided in https://docs.luxonis.com/projects/api/en/latest/components/nodes/yolo_detection_network/

jakaskerl

Hi jenanaputra
Looks like your model expects 640x640 input size, while you are sending in 416. You can solve this by setting (i assume setPreviewSize function was used originally) previewSize to (640, 640).

When you convert the model using luxonis tools, you should have received a json file which has the necessary mask and anchor information.

I'm not sure about the IMU problem. Are you trying to obtain IMU data anywhere in the code?

Thanks,
Jaka