PaddleSeg
PaddleSeg copied to clipboard
Easy-to-use image segmentation library with awesome pre-trained model zoo, supporting wide-range of practical tasks in Semantic Segmentation, Interactive Segmentation, Panoptic Segmentation, Image Mat...
### 问题确认 Search before asking - [X] 我已经搜索过问题,但是没有找到解答。I have searched the question and found no related answer. ### 请提出你的问题 Please ask your question How can I determine which model is...
### 问题确认 Search before asking - [X] 我已经搜索过问题,但是没有找到解答。I have searched the question and found no related answer. ### 请提出你的问题 Please ask your question 论文里面提到用超过1w张人像和7.7w张背景进行训练,训练结果很优秀,GitHub提供的human_matting-resnet34_vd.pdparams,与论文里面结果相似,但仍有差距,请问是否两者并非同一模型呢?
### 问题确认 Search before asking - [X] 我已经搜索过问题,但是没有找到解答。I have searched the question and found no related answer. ### 请提出你的问题 Please ask your question 问题一:我尝试直接使用demo却没有出现结果 问题二:我尝试训练出现trans_info为负值的错误 使用环境:linux 3090 python 3.9 配置环境:Package...
### 问题确认 Search before asking - [X] 我已经搜索过问题,但是没有找到解答。I have searched the question and found no related answer. ### 请提出你的问题 Please ask your question 您好,我用自己的数据集训练了一个MaskFormer的模型,参考https://github.com/PaddlePaddle/PaddleSeg/blob/release/2.9/docs/model_export_onnx_cn.md 可以将paddle的模型成功转换为预测模型文件。但是将预测模型文件转为ONNX时,提示  请问如何解决这个问题呢?谢谢。
### 问题确认 Search before asking - [X] 我已经搜索过问题,但是没有找到解答。I have searched the question and found no related answer. ### 请提出你的问题 Please ask your question 数据用的paddx stfpm里面范例的数据 python tools/train.py --config configs/stfpm/stfpm_resnet18_mvtec.yml 训练正常...
### 问题确认 Search before asking - [X] 我已经搜索过问题,但是没有找到解答。I have searched the question and found no related answer. ### 请提出你的问题 Please ask your question 请问是不是 只有 多尺度训练 的模型 在导出ONNX的时候才可以动态尺寸导出, 如果我训练网络固定了尺寸,比如输入的图片都是1280X720 这边做预处理都resize成640X360...
### 问题确认 Search before asking - [X] 我已经搜索过问题,但是没有找到解答。I have searched the question and found no related answer. ### 请提出你的问题 Please ask your question 需要用uint16的单通道灰度图像来训练,已经在Compose类里面把OpenCV的读入标志cv2.IMREAD_GRAYSCALE改成cv2.IMREAD_UNCHANGED了,config里也设置了img_channels=1,但得到的预测结果是没有意义的结果。想知道还有哪些地方需要修改的。PS:uint16直接转为tensor的bfloat16可能出现精度溢出问题,但我看框架会在预处理时先将图片转为float32,按道理应该不会是精度的问题
### PR types [Bug fixes] ### PR changes [Models] ### Description The Dice loss in `paddleseg.models.losses.dice_loss` and `paddleseg.models.losses.maskformer_loss` is modified based on [JDTLoss](https://github.com/zifuwanggg/JDTLosses/blob/master/losses/jdt_loss.py) and [segmentation_models.pytorch](https://github.com/qubvel-org/segmentation_models.pytorch/blob/main/segmentation_models_pytorch/losses/_functional.py). The original Dice loss is...
### 问题确认 Search before asking - [X] 我已经搜索过问题,但是没有找到解答。I have searched the question and found no related answer. ### 请提出你的问题 Please ask your question https://github.com/PaddlePaddle/PaddleSeg/tree/release/2.9.1/EISeg/med3d 文档里面的百度盘密码不对,麻烦重新提供一下。谢谢。