Error in demo/ocr_image_demo.py script

[narenderkumarnain]

I was trying the demo ocr_image_demo.py script and got this error.

[narenderkumarnain]

I was trying to use the Batch Inference feature added through a pr this morning, It worked well with the detection Model but Fails with the Recognition Model. Not even able to Use inference module with single image in batch (Just image Numpy array as input). The above fix just Suppresses the Batch Inference of Detection as well. So Recognition models are not yet able to support the Batch Inference or I am missing something? @cuhk-hbsun

[cuhk-hbsun]

@narenderkumarnain Thanks for your feedback.

1. Theoretically, for the algorithms using MultiRotateAugOCR in test_pipeline of text recognition or MultiScaleFlipAug in test_pipeline of text detection, the batch model is not supported since there may be multiple scaled images for one image.
2. But in practice, there is only one scale in most of the algorithms using MultiScaleFlipAug, or one rotate for MultiRotateAugOCR, so batch mode can be supported by turn off (ignore) the aug_test.
3. I will update the code to support batch mode for both text detection and recognition.

[cuhk-hbsun]

@narenderkumarnain I have updated the demo scripts. Please pull the latest code and try again.

[narenderkumarnain]

@cuhk-hbsun It works now. Thanks for the Changes.