MaskTextSpotter

MaskTextSpotter copied to clipboard

Recognition

2

Hello, is it possible to have the standalone recognition model ? Thank you.

AriouatI

module 'torch.distributed' has no attribute 'deprecated'

3

### pip list ### toolz 0.10.0 torch 1.2.0 torchvision 0.4.0a0 ### finetune.yaml ### SOLVER: IMS_PER_BATCH: 1 > python -m torch.distributed.launch --nproc_per_node=1 tools/train_net.py --config-file configs/finetune.yaml

LittleCurry

I found the model downloaded cannot reach the 0.86, 0.87, 0.87 for P, R, F for the icdar2015. Then I trained a model from the scratch, the code is not...

louviya

Can you provide the codes for evaluating the end-to-end recognition performance of the Total-text dataset? Thank you

Ineedangoodidea

It seems that the recognition result for long words is not good, and there are problems similar to attention drift. For example, 'rettungsweg' is recognized as 'retuungsweg'. How to solve...

ustczhouyu

how to get label of segmentations in SynthText

10

Thank you for your work. Could you please release the code of pre-processing of SynthText.

TriplePool

Hello, what is the meaning of GenericVocabulary_new.txt and GenericVocabulary_pair_list.txt? I can only find ch4_GenericVocabulary.txt in ICDAR website. What is the different betwenn them? lexicon_path = '../../lexicons/ic15/GenericVocabulary_new.txt' lexicon_fid=open(lexicon_path, 'r') pair_list =...

ustczhouyu

Sorry to bother you! In the file bounding_box.py, crop and rotate will change the chars' locations, but these operations keep the origin char_masks: if not isinstance(v, torch.Tensor) and k !=...

zhongdajian

Leojc

Leojc