求助！识别模型自训练后，使用训练模型识别准备，转为推测模型后识别出完全错误的文本内容

[zhangyistruggle]

使用的是rec_mv3_none_bilstm_ctc_v2.0_train预训练模型，配置yml是rec_chinese_lite_train_v2.0.yml。采用了自己的数据集进行识别模型训练。使用的是训练集图片对训练模型进行识别测试，脚本如下： python3 tools/infer_rec.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml -o Global.pretrained_model=./output/rec_chinese_lite_v2.0/best_accuracy Global.infer_img=./rec_test_image/image_1_crop_0.jpg 图片识别结果如下： 使用模型转换脚本： python3 tools/export_model.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml -o Global.pretrained_model=./output/rec_chinese_lite_v2.0/best_accuracy Global.save_inference_dir=./inference/rec_crnn 再次使用转换后的推测模型进行同一张图片的测试，脚本如下： python3 tools/infer/predict_rec.py --image_dir=./rec_test_image/image_1_crop_0.jpg --rec_model_dir=./inference/rec_crnn/ --use_gpu=false 图片结果识别如下：

模型转换前后识别结果差异很大。inference模型识别完全是错误的。 识别的图片如下：

[zhangyistruggle]

yml的参数如下： Global: use_gpu: false epoch_num: 500 log_smooth_window: 20 print_batch_step: 10 save_model_dir: ./output/rec_chinese_lite_v2.0 save_epoch_step: 3

evaluation is run every 5000 iterations after the 4000th iteration

eval_batch_step: [0, 1000] cal_metric_during_train: True pretrained_model: ./pretrain_models/rec_mv3_none_bilstm_ctc_v2.0_train/best_accuracy checkpoints: save_inference_dir: use_visualdl: False infer_img: doc/imgs_words/ch/word_1.jpg

for data or label process

character_dict_path: ppocr/utils/ppocr_keys_v1.txt character_type: ch max_text_length: 25 infer_mode: False use_space_char: True save_res_path: ./output/rec/predicts_chinese_lite_v2.0.txt

Optimizer: name: Adam beta1: 0.9 beta2: 0.999 lr: name: Cosine learning_rate: 0.001 regularizer: name: 'L2' factor: 0.00001

Architecture: model_type: rec algorithm: CRNN Transform: Backbone: name: MobileNetV3 scale: 0.5 model_name: small small_stride: [1, 2, 2, 2] Neck: name: SequenceEncoder encoder_type: rnn hidden_size: 48 Head: name: CTCHead fc_decay: 0.00001

Loss: name: CTCLoss

PostProcess: name: CTCLabelDecode

Metric: name: RecMetric main_indicator: acc

Train: dataset: name: SimpleDataSet data_dir: ./train_data/dazhongdianping_rec/train/ label_file_list: ["./train_data/dazhongdianping_rec/train/rec_train_gt.txt"] transforms:

• DecodeImage: # load image img_mode: BGR channel_first: False
• RecAug:
• CTCLabelEncode: # Class handling label
• RecResizeImg: image_shape: [3, 32, 320]
• KeepKeys: keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order loader: shuffle: True batch_size_per_card: 2 drop_last: True num_workers: 0

Eval: dataset: name: SimpleDataSet data_dir: ./train_data/dazhongdianping_rec/test/ label_file_list: ["./train_data/dazhongdianping_rec/test/rec_test_gt.txt"] transforms:

• DecodeImage: # load image img_mode: BGR channel_first: False
• CTCLabelEncode: # Class handling label
• RecResizeImg: image_shape: [3, 32, 320]
• KeepKeys: keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order loader: shuffle: False drop_last: False batch_size_per_card: 2 num_workers: 0

[LDOUBLEV]

你的图片没有变化吗？’收藏‘ 识别错了不太可能识别 ‘ 浙菜’ 这样的词汇；

刚用官方模型 重新导出模型测试了一次，导出模型没有问题；

[zhangyistruggle]

图片没有变化。路径是指向同一张图片的，该图片是训练集的图片。

[LDOUBLEV]

• 再检查下测试数据吧
• 保留你现在的环境，重新clone PaddleOCR的代码重新测试看下

[Fzz123]

感觉应该是配置文件的问题。rec_mv3_none_bilstm_ctc_v2.0_train这个预训练模型所对应的配置文件应该不是轻量级模型的rec_chinese_lite_train_v2.0.yml。可以查看文件夹里的train.log中记录的配置信息来进行训练。

[paddle-bot-old[bot]]

Since you haven\'t replied for more than 3 months, we have closed this issue/pr. If the problem is not solved or there is a follow-up one, please reopen it at any time and we will continue to follow up. It is recommended to pull and try the latest code first. 由于您超过三个月未回复，我们将关闭这个issue/pr。 若问题未解决或有后续问题，请随时重新打开（建议先拉取最新代码进行尝试），我们会继续跟进。