danfenghong / ieee_tip_uiu-net Goto Github PK

Hello. How can I modify the input shape of the model?

你好，因为test输出的结果中平均值只有零点零几，所以我发现了几个问题：

   if IoU > best_iou:
        best_iou = IoU
    if nIoU > best_nIoU:
        best_nIoU = nIoU

    total_iou = total_iou + IoU
    total_niou = total_niou + nIoU`

这一部分代码是不是应该放到for循环里呀。
还有

IoU=total_iou/20
nIoU=total_niou/20

用新的名字结果就对了:

avg_IoU=total_iou/20
avg_nIoU=total_niou/20

测试结果iou比niou还小，是哪里出错了吗

作者尝试过把这个网络添加到yolov7中去么

您好，我在计算红外小目标检测传统算法IPI的时候遇到了问题。
我利用混淆矩阵计算指标，发现计算出来的IoU与nIoU是一样。因为IoU=tp/(fn+fp+tp)，nIoU=tp/(t+p-tp)。公式表示是一样的，我感觉我算出来的是nIoU。
请问您可以教教我，如何计算红外小目标检测传统算法的IoU与nIoU吗？

U^2 Net: https://github.com/xuebinqin/U-2-Net

请问论文中所述Synthetic dataset 选取6200张图像，具体是怎么实现的？有开源的代码吗？文中只说了手动去除预处理后具有零值或负值的图像，那具体是保留了哪些图像呢？

Can you please provide tutorial for training and testing the model on custom dataset

您好，我在使用其它两种数据集训练时，在一个epoch下，不同批次会出现报错
l0: 0.010986, l1: 0.011472, l2: 0.014452, l3: 0.004034, l4: 0.004883, l5: 0.013490, l6: 0.004046

[epoch: 1/500, batch: 45/ 213, ite: 15] train loss: 0.884871, tar: 0.183994
Traceback (most recent call last):
File "D:\contrast-methods\IEEE_TIP_UIU-Net-main\train.py", line 139, in
for i, data in enumerate(salobj_dataloader):
File "C:\Users\hp-\AppData\Roaming\Python\Python311\site-packages\torch\utils\data\dataloader.py", line 630, in next
data = self._next_data()
^^^^^^^^^^^^^^^^^
File "C:\Users\hp-\AppData\Roaming\Python\Python311\site-packages\torch\utils\data\dataloader.py", line 1345, in _next_data
return self._process_data(data)
^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\hp-\AppData\Roaming\Python\Python311\site-packages\torch\utils\data\dataloader.py", line 1371, in _process_data
data.reraise()
File "C:\Users\hp-\AppData\Roaming\Python\Python311\site-packages\torch_utils.py", line 694, in reraise
raise exception
RuntimeError: Caught RuntimeError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "C:\Users\hp-\AppData\Roaming\Python\Python311\site-packages\torch\utils\data_utils\worker.py", line 308, in _worker_loop
data = fetcher.fetch(index)
^^^^^^^^^^^^^^^^^^^^
File "C:\Users\hp-\AppData\Roaming\Python\Python311\site-packages\torch\utils\data_utils\fetch.py", line 54, in fetch
return self.collate_fn(data)
^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\hp-\AppData\Roaming\Python\Python311\site-packages\torch\utils\data_utils\collate.py", line 265, in default_collate
return collate(batch, collate_fn_map=default_collate_fn_map)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\hp-\AppData\Roaming\Python\Python311\site-packages\torch\utils\data_utils\collate.py", line 127, in collate
return elem_type({key: collate([d[key] for d in batch], collate_fn_map=collate_fn_map) for key in elem})
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\hp-\AppData\Roaming\Python\Python311\site-packages\torch\utils\data_utils\collate.py", line 127, in
return elem_type({key: collate([d[key] for d in batch], collate_fn_map=collate_fn_map) for key in elem})
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\hp-\AppData\Roaming\Python\Python311\site-packages\torch\utils\data_utils\collate.py", line 119, in collate
return collate_fn_map[elem_type](batch, collate_fn_map=collate_fn_map)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\hp-\AppData\Roaming\Python\Python311\site-packages\torch\utils\data_utils\collate.py", line 162, in collate_tensor_fn
return torch.stack(batch, 0, out=out)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: torch.cat(): input types can't be cast to the desired output type Byte。
请问这种错误有解决方案吗？

你好，提供的train_data是空的，但是train.py包含如下：
tra_image_dir = os.path.join('DUTS', 'DUTS-TR', 'DUTS-TR', 'im_aug' + os.sep)
tra_label_dir = os.path.join('DUTS', 'DUTS-TR', 'DUTS-TR', 'gt_aug' + os.sep)
我应该从哪里下载'DUTS', 'DUTS-TR'，'im_aug' 'gt_aug' 这些

i can't download the model because i should have an account in baidu and this app works for chinese

i want to ask if there are any precise versions of the requirements (specially for torch, torchvision) because it didn't work for me without specifying it . I'm looking for a quick response please. and if you can help me with any information about the metrics because the way you coded  IoU and nIoU have a really small value0.000xx even if there is an intersection (visually noticed) . I tried to print the intersection to check, and it is = 0 and the comparison of the label and predicted mask shows the opposite (a good result).

代码中iou的计算感觉有点问题。

您好，我这里尝试了多次配置uiunet的相关环境，但requirements的部分依赖包版本未指明，导致都未能成功，我也只是一个初学者，但我个人感觉确实这个环境的配置成功的难度较大，所以提一个小建议

利用网盘的uiunet.pth测试代码的数据，实验指标分别为
Quantification_Results
0.00200256328099968 0.004197722567287785
0.040051265619993594 0.0839544513457557
Verification_data
0.0 0.0
0.0