Hi @Bumbleboss
Try to convert the model to .onnx and then to .blob with https://blobconverter.luxonis.com/, but usually with the Yolo type models, we are pruning the detection head (the tail), because in depthai there's the on-device postprocessing implemented that we use, therefore we are pruning the post-processing in the model weights during the export in the luxonis tools.
Thanks,
Jaka