Why crop_orientation_predictor so slow? #1930
Replies: 2 comments
-
P.S. I tried the crop_orientation_predictor from the onnxtr library but it is even slower |
Beta Was this translation helpful? Give feedback.
-
Hi @decadance-dance 👋🏼, You are right that comes from the resolution the models have to process. recognition: 32x128 Inputs Mh but it should be faster or at a minimum comparable between docTR and OnnxTR do you run OnnxTR also on CUDA ? |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Bug description
I use the crop_orientation_predictor together with the recognition_predictor in my code and I noticed that it is kinda slow. The crop_orientation_predictor inference speed is approximately equal to the recognition_predictor one.
I expected it to be pretty fast since it is a small model.
Code snippet to reproduce the bug
Error traceback
Environment
Deep Learning backend
is_tf_available: False
is_torch_available: True
Beta Was this translation helpful? Give feedback.
All reactions