Tingquan Gao

Results 71 comments of Tingquan Gao

fix https://github.com/PaddlePaddle/PASSL/issues/175

我看了下文档,测试了下。`paddle.to_tensor(value)`得到的是一个0-D Tensor,所以是否应该使用`paddle.full(shape=[], value)`这样?

如果文物的种类固定,并且训练集固定,建议使用目标检测完成。如果文物种类不固定(需要经常变化、调整),建议使用PP-ShiTu。

修改数据集路径(image_root: ./dataset/ cls_label_path: ./dataset/train_reg_all_data_v2.txt)后可以先训练试试,观察loss是否下降,已经最终的收敛情况、精度情况,再适当调整learning rate。

收到,我们排查一下该问题,尽快回复。

不好意思,因为人员属性模型是多标签分类模型,目前paddle serving还不支持多标签。我们后期会考虑支持该功能,谢谢关注!

还有车辆属性是多标签分类: https://github.com/PaddlePaddle/PaddleClas/blob/release%2F2.5/docs/zh_CN/models/PULC/PULC_vehicle_attribute.md

thx for your reply. i want to reproduce this model, so it is best if the pretrained could be provided. i would retrain using the code.

hi, because want to reproduce this work, i tried to train the resnet18 using [image_classification_sota](https://github.com/hunto/image_classification_sota/) and the top1 acc got is 70.8 that is better than proposed in the paper(69.54)....

做服务化部署的话,建议使用PaddleX导出服务化部署包,或是了解一下PaddleServing。