Replies: 1 comment
-
Researching more about this bug, it seems that there is a bug from the original onnx_export model in the Rapid repo, because the problem from that model is the same as the final tflite converted model. Maybe you can check a little bit that export_onnx.py script, because i can't find why the final output is so different, maybe is because some image resizing. Thanks a lot |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi!
Im testing the tflite converted model and comparing with the original pretrained one (using the float32 model), but the output is quite different from one to the other, is the any conversion that you made that I'm missing?
Im just using the PIL images transposed and converted to numpy arrays.
Thanks a lot
Beta Was this translation helpful? Give feedback.
All reactions