gengyanlei / segmentation_pytorch Goto Github PK

我的数据集size不一样，分辨率很大，是一个二分类问题

loss回传时突然从tensor转成地址了

我用了您的代码跑了您的data里的几张图片 发现会在for step, data in enumerate(data_loader['train'])这步报错，原因是
Traceback (most recent call last):
File "D:/cc/segmentation_pytorch-master/main.py", line 80, in
main(args)
File "D:/cc/segmentation_pytorch-master/main.py", line 66, in main
train(data_loader, model, optimizer, scheduler, tb_writer, param_dict, continue_epoch)
File "D:\cc\segmentation_pytorch-master\utils\trainval.py", line 32, in train
for step, data in enumerate(data_loader['train']):
File "D:\ProgramData\Anaconda3\lib\site-packages\torch\utils\data\dataloader.py", line 363, in next
data = self._next_data()
File "D:\ProgramData\Anaconda3\lib\site-packages\torch\utils\data\dataloader.py", line 403, in _next_data
data = self._dataset_fetcher.fetch(index) # may raise StopIteration
File "D:\ProgramData\Anaconda3\lib\site-packages\torch\utils\data_utils\fetch.py", line 44, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "D:\ProgramData\Anaconda3\lib\site-packages\torch\utils\data_utils\fetch.py", line 44, in
data = [self.dataset[idx] for idx in possibly_batched_index]
File "D:\cc\segmentation_pytorch-master\utils\dataset.py", line 51, in getitem
image, label = self.transform(image, label)
File "D:\cc\segmentation_pytorch-master\utils\aug_PIL.py", line 228, in call
image, label = t(image, label)
File "D:\cc\segmentation_pytorch-master\utils\aug_PIL.py", line 181, in call
image, label = getattr(self.aug_pil, aug_name)(image, label)
File "D:\cc\segmentation_pytorch-master\utils\aug_PIL.py", line 65, in random_resize_crop
image = tf.resized_crop(image, i, j, h, w, self.input_hw, interpolation=Image.BILINEAR)
File "D:\ProgramData\Anaconda3\lib\site-packages\torchvision\transforms\functional.py", line 499, in resized_crop
img = resize(img, size, interpolation)
File "D:\ProgramData\Anaconda3\lib\site-packages\torchvision\transforms\functional.py", line 324, in resize
raise TypeError('Got inappropriate size arg: {}'.format(size))
TypeError: Got inappropriate size arg: (256, 256)
这是什么原因 一直没解决

从yaml文件里面读出来的str不转换成float的话，学习率那就有个报错

可以跑起来吗？咋没有结果呢？

老哥，你遇到过这个问题吗，我看了一下rotate这个函数确实是有fill参数的，为啥还会报这个错呢

你好，针对deeplab_v3_50.py，我发现你修改的网络普遍不加BN，但是效果也不差。请问你有什么特殊的考虑吗？望指教，谢谢。

Hi, I implemented a pytorch version for deeplabv3+ (it is unorganized still). I only achieved 65% in PASCAL2012 using the augmented PASCAL in the training phase. This happened due to I didn't use the complete Xception-like backbone. I think you should try to read the backbone (or maybe the complete network in order to compare the models) from the TF model before training. This GitHub describes a code to read the model from TF for Keras.

https://github.com/bonlime/keras-deeplab-v3-plus/blob/master/extract_weights.py

作者您好：
阅读了你的代码，很受启发，有一个问题需要请教一下：
lab=self.label[index].argmax(axis=-1) # no one-hot
为什么需要argmax，在label中对应正样本像素为1，负样本为0，那么通过argmax后，如果label是二维图像，那么argmax导致降维，这如何同网络输出对应的

像人脸多任务合并，onnx2tensorRt等多个项目都在更新，代码编写比较慢。