Question/Issue:
After the deployment of the object detection project (91.9% validation accuracy and 94% test accuracy) to Raspberry Pi 4 using Linux Python SDK and Linux EIM executable, the accuracy seemed to reduce drastically. I tried to run the models locally and through the edge impulse browser and they have shown very different results. The signal processing blocks are claimed to be included in the impulse within the EIM file so I don’t know where things are going wrong. The model is selected from results of the EON Tuner. Thank you for the help!
Project ID:
357241
Context/Use case:
1 Like
the classify-image.py from (https://github.com/edgeimpulse/linux-sdk-python/tree/master/examples/image) works for me when i resize the input image to the model. thank you!
1 Like