Saleem Ahmed

@cjnolet I have a similar use case, please advise : 1. elaborate on how to extract binary masks, can I generate them per character or per word. 2. How to...

Same error. There is no file routing.py in chatbot_website/chatbot_website So I changed line 73 of chatbot_website/chatbot_website/settings.py from `ASGI_APPLICATION = 'chatbot_website.routing.application'` to `ASGI_APPLICATION = 'chatbot_interface.routing.application'` But then I realised [django channels...

run the docker image, that worked for me. or maybe port code to channels 2.

https://github.com/crazysal/SeqGAN/ Needed some minor changes for porting to py 3

> the reason is the pytorch and cuda version, it works well now. which version does it work with ??

mask_gt is generated only for dataset having character level annotation : Synthtext. Check section 2.3 of paper for training strategy. mask_iou_angle is generated from output of East proposals in case...

You will need to retrain for new dataset. Instead of training from scratch could try to fine-tune this model from the weights published in google drive ..

Training only on Synthtext with fixed trained east base(benchmark matched ) and groundtruth bounding boxes from Synthtext gives recognition accuracy plateaued at 20%

Update : Using pre-trained glove embedding (https://github.com/minimaxir/char-embeddings) I am getting 56.7% accuracy on SynthText now.. Any comments on results on SynthText would be appreciated.

sample_gt_label_input is one hot vector embedding of char label from groundtruth concatenated to first attention Lstm output check here : for model visualisation http://ethereon.github.io/netscope/#/gist/edc52a3554d5f1e8ee6a24d774de2bd0