Sorry, I forgot to mention but I rolled back to 4.19 and the same error still persists. I’m also using localhost before I accidentally updated my raspberry pi. Also, can I dm my code to you? Just in case, if it would be of any help
hi @shashi , do you know how I can use this “yolov8n_relu6_lp_ocr–256x128_quant_hailort_hailo8l_1.hef” without degirum_tools? I’m trying to make a script to detect and recognize license plate with a custom model for detection and this model for OCR similar to this examples - Hailo-Application-Code-Examples/runtime/python/object_detection/object_detection.py at main · hailo-ai/Hailo-Application-Code-Examples · GitHub
I was able to run the detection and it works correct, but then when I’m trying to run OCR, then the result of inference is 6 outputs with arrays of numbers.
I do not understand how to process these numbers, do you know what could be the starting point?
Thanks!
Hi @Aleksei_Markov
If you want to use our models outside of PySDK, you need to write your own postprocessor that takes the 6 output tensors.
Do you know if it’s possible to see how you process these 6 output tensors using degirum?
@Aleksei_Markov
Is there a reason why you do not want to use PySDK?
I want to have a custom solution as much as we can
@Aleksei_Markov
Thanks. In that case, you need to write your own post-processing code. You can study the code in ultralytics repo to understand how output tensors are processed to decode bboxes and class probabilities and finally converted to human readable labels. Since our code is in C++ and deeply integrated into our library, we cannot share the code. However, the logic used in quite well known so there is no secret recipe inside. If you follow ultralytics code, you should be able to replicate.
I got you. Thanks!
btw, do you know what part of ultralytics code is responsible for it?
Oh, I just realized that you are co-founder of Degirum, now it makes sense for me. I thought that you are from Hailo team
Hi @Aleksei_Markov
You can look at the inference and postprocess functions in Detect class for the logic: ultralytics/ultralytics/nn/modules/head.py at main · ultralytics/ultralytics