search

PFCCLab / StyleText

22 stars 0 forks source link

合成的图片不遵循具体的输入文本，感觉有点像乱码 #1

Open jaycecd opened 1 month ago

jaycecd commented 1 month ago

你好，发现Windows平台使用当前版本的StyleText，依照使用教程中给出的指令： python3 tools/synth_image.py -c configs/config.yml --style_image examples/style_images/2.jpg --text_corpus PaddleOCR --language en 生成的图片几乎是不可用的： fake_text： fake_text fake_fusion： fake_fusion 但使用之前集成在paddleOCR（2.5.0.3，cuda 10.2, cudnn 7.6）中的StyleText能够正常生成。 请问这是什么问题？ (pdocr_2.5.0.3) D:\StyleText-new>python tools/synth_image.py -c configs/config.yml --style_image examples/style_images/2.jpg --text_corpus PaddleOCR --language en [2024/09/23 18:49:52] srnet INFO: load pretrained model from style_text_models/bg_generator [2024/09/23 18:49:54] srnet INFO: load pretrained model from style_text_models/text_generator [2024/09/23 18:49:54] srnet INFO: load pretrained model from style_text_models/fusion_generator

Chevolier commented 1 month ago

Same issue, but in ubuntu system, with paddleocr=2.8.1, any solutions?

GreatV commented 1 month ago

需要debug一下问题