Text_Detector
Text_Detector copied to clipboard
Text detection model that combines Retinanet with textboxes++ for OCR
Hi, I need to convert ./ .index .meta data to ./ .pb variable/ .index .data but I could not figure out a way to do so please can anyone help...
I use the model downloaded form the link:ICDAR2015_TextBoxes.pth, but only got hmean = 77.38% in inference python eval.py --input_size=1280 --nms_thresh=0.2 --cls_thresh=0.4 --dataset=ICDAR2015
Where do we specifically get these weights/v1_step1/ckpt_3000.pth?
In the Detector directory there is no Retinanet 'from Detector.RetinaNet import RetinaNet' therefore fails. Should I get it from somewhere?
I have my own dataset. Can I use this model for a different dataset apart from the ones in the paper? Thanks
can you provide the loss graph during training? my TRAIN_TOTAL_LOSS is always round 1.2
could you give the ImageNet pre-trained weights link, thanks.
Hello, how can I change the project to support multiple classification of text box detection tasks?
Hi, I'm curious what is your step 1/2/3? I notice that they are using different datasets, so is that mean if I want to train I must run the step...
thanks for your project. I'm wondering why, when computing IOU, compute the IOU of a rectangle, not a polygon