search

ying09 / TextFuseNet

A PyTorch implementation of "TextFuseNet: Scene Text Detection with Richer Fused Features".
MIT License
474 stars 121 forks source link

SynthText pretrained model #70

Closed jianpingZhonggit closed 2 years ago

jianpingZhonggit commented 3 years ago

Can you provide the pretrained weight on SynthText? There only offer bouding box annotations of word,character in SynthText? However the TextfuseNet needs segmentation annotation,can you show the code how to get th segmentation annotations? Thank you

Real-YeJ commented 3 years ago

@jianpingZhonggit hey, the pretrained-model on SynthText can be download on Google driver. For Synthtext, you can use the bouding box annotations as the segmentation annotations. For your custom datasets, you can use our pretrained-model to generated pseudo labels.

jianpingZhonggit commented 3 years ago

Oh, Thany you very much!Using the bouding box as segmentation in character is acceptable.However,can we use the bouding box of the word as segmentation annotation? That means bounding box equals the segmentation

jianpingZhonggit commented 3 years ago

@Real-YeJ Oh, Thany you very much!Using the bouding box as segmentation in character is acceptable.However,can we use the bouding box of the word as segmentation annotation? That means bounding box equals the segmentation.

Real-YeJ commented 3 years ago

@jianpingZhonggit horizontal texts √, multi-oriented and arbitrary-shaped texts ×.

jianpingZhonggit commented 3 years ago

@Real-YeJ Thank you very much! I will try again.