PaddleSeg icon indicating copy to clipboard operation
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...

Results 373 PaddleSeg issues
Sort by recently updated
recently updated
newest added

Thanks for your issue. To help us solve the issue better, please provide following information: 1. PaddleSeg version: (please specify the branch as well,e.g. PaddleSeg release/2.3) 2. PaddlePaddle version: (e.g....

question

While optimizing PP-HumanMatting model I have found that the computational graph contains some weird and non-optimal pattern. Instead of using `pad2d` op, there is a combination of `unsqueeze2` + `pad3d`...

question

from paddleseg.models import PPLiteSeg from paddleseg.models.backbones import STDC2 model = PPLiteSeg(num_classes=2, backbone=STDC2) 执行上述命令,会出现 AssertionError: The backbone should has feat_channels.

question

Thanks for your issue. To help us solve the issue better, please provide following information: 1. PaddleSeg version: (please specify the branch as well,e.g. PaddleSeg release/2.3) 2. PaddlePaddle version: (e.g....

question

我使用python -m paddle.distributed.launch train.py --config configs/pspnet/pspnet_resnet101_os8_cityscapes_1024x512_80k.yml --do_eval --save_dir /mnt/Workdir/pspnet/pspnet_resnet101_os8_cityscapes_1024x512_80k/ --num_workers 12 为例 为什么 时长会到达40小时 2022-06-28 07:04:06 [INFO] [TRAIN] epoch: 1, iter: 20/80000, loss: 2.1961, lr: 0.009998, batch_cost: 1.7653, reader_cost: 0.00016,...

question

Hi Thanks for sharing this project. I want to train a new model for **lane segmentation** and inference it on **Raspberry Pi**. What do you think is the best model...

您好!我在使用PaddleViT设计自己的模型,对于实验的结果我遇到了一些问题,希望可以得到帮助。首先,因为PaddleViT和PaddleSeg计算评价指标的代码我都仔细看了,是一样的。可以查阅https://github.com/BRIDL/PaddleViT/blob/develop/semantic_segmentation/src/utils/metrics.py, 以下图片是我在数据集上的实验结果,框起来的是我用PaddleViT平台跑出来的实验结果。其余的FCN、PSPNet等模型是我自己的代码的结果。 ![image](https://user-images.githubusercontent.com/40911688/175894733-d9dbd021-194e-4bb6-a9ee-9fccf01cf605.png) OA和Miou的计算公式是: ![image](https://user-images.githubusercontent.com/40911688/175895809-c39411e2-161b-4be6-b51b-80ee97e51d94.png) ![image](https://user-images.githubusercontent.com/40911688/175895849-28f52bc7-08f2-458b-a4af-52008b61f238.png) 按理说mIOU值大的OA也应该高才对,从我的实验结果在的确是mIOU达到了最高,但是OA的指标是最低的,这很明显有问题 问题1:paddleseg/utils/metrics.py中的函数:def class_measurement: --- mean_acc是不是论文中常见的OA指标?如果不是,应该怎样通过三个参数intersect_area, pred_area, label_area实现呢 ![image](https://user-images.githubusercontent.com/40911688/175897626-087abda4-7ab2-45b0-977f-c3be127145e7.png) ![image](https://user-images.githubusercontent.com/40911688/175896725-1860a8c4-6957-41ce-8883-222fc50b92f3.png) 问题2:能简单解释一下是Paddle怎么计算评价指标的吗?代码显示mIOU, OA都没有根据混淆矩阵计算,而是使用的三个参数: intersect_area, pred_area, label_area进行计算,对应于混淆矩阵应该怎样理解这三个参数呢,如intersect_area就是混淆矩阵对角线素元素和np.diag(hist),其他的 hist.sum(axis=1) ,hist.sum(axis=0)又是怎么对应的呢? 问题3:以下是我在验证过程中使用的代码,是我的计算方式出问题了吗?上图中自己实现的一些模型FCN,PSPNet(没有使用paddle平台)等模型虽然mIOU的很低,但是mIOU相对较高?而我使用paddleViT运行的模型mIOU都达到了72.99,OA却只有85.73的情况?我想整合这两个评价指标。 在验证过程中调用评价指标函数 ![image](https://user-images.githubusercontent.com/40911688/175902317-e3969c4b-5ede-4a51-86ab-3bd26173e6be.png) 评价函数具体实现:通过混淆矩阵计算 ![image](https://user-images.githubusercontent.com/40911688/175901659-c46f522d-9b79-43c5-9078-d629d85ca33b.png) ![image](https://user-images.githubusercontent.com/40911688/175901732-8654f9e9-80ec-415b-a5f5-c05dabfc1000.png) ![image](https://user-images.githubusercontent.com/40911688/175901808-5047d2e3-9662-4883-8fec-e2402df0fde8.png) ![image](https://user-images.githubusercontent.com/40911688/175901831-353b67b2-278b-4385-9a95-83016e0a43b2.png)...

question

1. PaddleSeg version: PaddleSeg v2.5.0 (release/2.5) 2. PaddlePaddle version: 2.3.0 3. Operation system: Linux-5.13.0-51-generic-x86_64-with-glibc2.17 4. Python version: 3.8.10 5. CUDA/cuDNN version: NVCC: Build cuda_11.6.r11.6/ cudnn: 8.4 6. Problem: I want...

question