YF

移动字/2021-12-10 013543.bmp [{"transcription": "茅", "points": [[260, 121], [335, 119], [336, 173], [261, 175]], "difficult": false}, {"transcription": "笔", "points": [[369, 133], [428, 133], [428, 190], [369, 190]], "difficult": false}, {"transcription": "或",...

试过用示例的格式，单张图片训练单个字或者句子可以成功

有没有办法让识别只返回单个字而不是整个句子

明白了

 比如图片上这样，我是根据这样来选择对应的预训练模型的，这样应该没错吧 ch_PP-OCRv3_det_cml.yml 使用 ch_PP-OCRv3_det_distill_train，训练ch_PP-OCRv3_det训练2600次预测出来的precision:0.3717948717948718 det_mv3_db.yml 使用 MobileNetV3_large_x0_5_pretrained.pdparams，训练5800次预测出来的precision:0 这让我产生了怀疑，也许就是预训练权重不对应导致的 请问文件标题文字相似度高权重文件的就是吗

Guys, I have this problem too, how do I solve it now, virtual memory is loaded before each iteration, memory is released after completion of iteration, memory is loaded again,...

Now I use the command to temporarily solve this problem, yolo detect train data=coco128.yaml model=yolov8n.pt epochs=100 workers=0 name=test batch=20 cache=FALSE