DocTr++论文复现 #11475
DocTr++论文复现 #11475
Conversation
|
Thanks for your contribution! |
|
chenjiajun05 seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account. You have signed the CLA already but the status is still pending? Let us recheck it. |
There was a problem hiding this comment.
非常感谢您的贡献~ 请提交训练参数和inference模型方便我们进行验证。 另外辛苦将代码融合进PaddleOCR框架，可以参考:
https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.7/applications/%E9%AB%98%E7%B2%BE%E5%BA%A6%E4%B8%AD%E6%96%87%E8%AF%86%E5%88%AB%E6%A8%A1%E5%9E%8B.md
| image_path = os.path.join(self.root, "img", image_name + ".png") | ||
| image = cv2.imread(image_path) | ||
|
|
||
| # if image is None: |
| echo " # done" | ||
| } | ||
|
|
||
| doc3d_download "http://vision.cs.stonybrook.edu/~sagnik/doc3d/img_1.zip" "$outputPath/" "img_1.zip" |
applications/Doctr++文档矫正/train.sh
Outdated
| export FLAGS_logtostderr=0 | ||
| export CUDA_VISIBLE_DEVICES=4 | ||
|
|
||
| python train.py --data-root /ssd1/chenjiajun05/chenjiajun05/doc3d \ |
There was a problem hiding this comment.
不要包含个人目录，建议删除sh脚本，将启动训练说明写在文档中
| ./train.sh | ||
| ``` | ||
| 每个époch后，会进行验证集评估，保存best_model以及last_model。 | ||
| ## 7. 模型评估 |
| chmod +x train.sh | ||
| ./train.sh | ||
| ``` | ||
| 每个époch后，会进行验证集评估，保存best_model以及last_model。 |
| 准备工作完成后，即可开始进行模型训练。本项目提供训练脚本train.sh。其中需要自行修改数据集路径--data-root | ||
| ```python | ||
| # 模型训练 | ||
| chmod +x train.sh |
There was a problem hiding this comment.
将脚本替换为完整启动命令，参考：
python3 tools/train.py -c configs/rec/PP-OCRv3/en_PP-OCRv3_rec.yml -o Global.checkpoints=./your/trained/model
| python eval_DocUNet.py --i ./crop/ --m pretrained_model --o output_val --g ./scan/ | ||
| ``` | ||
| 效果如下： | ||
|
|
There was a problem hiding this comment.
请增加推理部署相关说明：包括单图推理脚本，inference模型导出命令，以及基于inference模型预测脚本。参考：
https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.7/doc/doc_ch/algorithm_rec_nrtr.md#4-%E6%8E%A8%E7%90%86%E9%83%A8%E7%BD%B2
|
这个有训练好的模型吗？怎么使用？ |
DocTr++论文复现