-
Notifications
You must be signed in to change notification settings - Fork 45
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
How to do only text detection on minidocr? #691
Comments
Hello, for only detection, you could run python tools/infer/text/predict_det.py --image_dir {path_to_img or dir_to_imgs} --det_algorithm DB++ please refer to https://github.com/mindspore-lab/mindocr/blob/main/tools/infer/text/README.md for more details. In the stats, the F-score and throughput are available on 8xAscend 910 with mindspore 2.0. And if the detection and recognition is slow, you could try
You could also modify the items |
How do I do only text detection on minidocr using models from here:
https://github.com/mindspore-lab/mindocr/blob/main/configs/det/dbnet/README.md
-> section 3. Results -> DBNet and DBNet++ were trained on the ICDAR2015, MSRA-TD500, SCUT-CTW1500, Total-Text, and MLT2017 datasets ...
Let for example I want text detection speed and accuracy of this one on my dataset (images folder):
MLT2017 | DBNet | D910x8-MS2.0-G | ResNet-18 | SynthText that has throughput = 344.8 img/s
Any example any help is appreciated.
BTW: What does "Throughput" mean exactly? In this contest, I mean in terms of what machine those stats are available.
EDITED: 1. Also why is the box detection per each word individual instead of being the same bounding box when all words are in the same line as it is on paddleocr? Is there any easy way I can archive it?
2. This code is working to detect and recognize but is very slow (0.9 seconds to detect a single image 1152x104px on macbook pro m3) I need a faster one with lower accuracy trade off and separated detection from recognition:
The text was updated successfully, but these errors were encountered: