Youngmin Baek

@ntdat017 A silly answer to a wise question. The formula is based on empirical observation. :) ``` niter = int(math.sqrt(size * min(w, h) / (w * h)) * 2) ```...

Thanks, @mayank-git-hub for your explanation of Gaussian heatmap. Some comment about SynthText dataset is that some trascription are incorrectly labeled in SynthText. It does not guarantee that a word box...

@mayank-git-hub Oh, your code already has that functionality. I missed it. Please forget about my previous comment. :)

@thaihaquang Sorry for the late reply. Special characters like periods are hard to deal with. We used the first way by following the character bounding boxes in SynthText dataset. Also,...

@backtime92 1、what is the size of input image? - All images for training were resized to 768x768. An augmentation that enlarges the image using random crop is applied. 2、Did you...

@YanShuang17 In my experiments, the initial lr is 1e-4, and multiply 0.8 for every 10k iterations. However, in my opinion, the lr parameter was not sensitive when training. Rather, please...

@Godricly I have tried image-level forwarding too. Interesting you came up with the same idea as mine! I found that it will be a good option to use both of...

@sonack I did not use synchronized-bn for this work since the operator was not supported in pytorch at that time. Recently pytorch provide the official functionality of the SyncBatchNorm, and...

In CPU mode, it takes a long time, but I have no idea why the inference speed is slow in GPU mode. I think there are so many factors such...

I just followed the instruction provided by opencv document (https://docs.opencv.org/3.3.1/d3/db4/tutorial_py_watershed.html). We used thresholding for the binary maps for finding three areas such as sure_fg, sure_bg, and unknown in the example....