Awesome multilingual OCR toolkits based on PaddlePaddle (practical ultra lightweight OCR system, support 80+ languages recognition, provide data annotation and synthesis tools, support training and deployment among server, mobile, embedded and IoT devices)
aistudio 4卡gpu环境,paddleocr 2.6.0 中文古语识别,竖排文字,所以dataset里面的w、h我调换了顺序,设置成了[3, 640, 32] , reconaug设置了[640, 32, 3] 文本最长50字,已经根据gt切成了小图 启动命令
详细配置
检查图片
这个问题可能由什么引起?4卡128g都爆了