khornlund / severstal-steel-defect-detection Goto Github PK

After installing the segmentation-models-pytorch, the package is not found in the environment's site-packages. Even the sever package is not present. Installing segmentation-models-pytorch separately gives the package but compatibility errors with other packges arise while installing separately.
How we can sort the issue?

在机器视觉领域，**一直被halcon，康耐视和基恩士垄断。良心的我自主研发机器视觉自动化软件。在缺损检测领域有着独到的经验。
目前我的软件优势：
1、定位技术上不输halcon。
2、专利检测算法pww特征提取。可以将颜色纹理量化后提取区域轮廓计算量化的面积。
3、图像制程采用多层次定位+pww特征提取检测。比深度学习更可靠。
4、采用流程图和决策图的全中文运动制程。比plc更简单。
5、保留着halcon接口。支持halcon工程师的二次开发使用。

https://download.csdn.net/download/pww71/85093101
https://download.csdn.net/download/pww71/62047145
链接：https://pan.baidu.com/s/1vsTptn_pvtbK2sDhWVCZJg
提取码：1234

当前市场上很多类似软件和我的比差距很大 。首先他们的功能过于庞大，而且不够通用。学习和操作不是普通人能短时间掌握的。而我的软件优势明显。 就是定位和检测。其他的任何算子不论是halcon还是其他厂商的算子都可以定制。从外部接口导入到框架内。定位和检测都是自主研发，检测直接量化颜色纹理和区域轮廓进行分析，是我申报专利的算法。因此参数固定和简单。当然比深度学习参数还是麻烦一点点。但是效果比深度学习更稳定。
一般情况下，人眼识别都是颜色纹理和区域轮廓这些基本特征。所以人眼能识别的，基本上我的检测就能识别。而且定位采用多层次定位，在固定位的颜色纹理和区域轮廓上分析基本上可以满足市场上百分之90的缺损检测需求。
另外 我的框架是仿照操作系统的架构，支持任何软件和硬件，只要按照我的接口标准写驱动就可以融入框架。
所以 对于任何高速电机和板卡，3d相机，激光检测设备等等。 都可以融入我的框架。我的内部就只负责客户制程和指挥调度各个模块。
现在操作系统对各个硬件软件的支持也是通过开放的接口。我也是这样做的，对于高级开发应用还是需要定制的。而对于大量通用的检测。直接可以让工人制程就可以完成。

Hello，
when I run "sever train -c experiments/unet-b5.yml", there is an error:
TypeError: init() got an unexpected keyword argument 'dropout'
TypeError: init() got an unexpected keyword argument 'weight_std'
Then I delete keywords and it is able to run successfully, Why does this happen, have you deleted some modules ？

hello ,how can i training your model?

can you share me the weight file after training?

I've successfully started training but when I tried resuming from the last epoch it is starting the training from epoch 0.

(steel) C:\Users\User\Documents\sdd>sever train --resume "C:\Users\User\Documents\sdd\saved\sever-FPN-efficientnet-b5-BCEDiceLoss-RAdam\1008_124332\checkpoints\checkpoint-epoch179.pth"
5740 - Runner - INFO - Using random seed: 447676
5740 - Runner - DEBUG - Building model architecture
Load result: None
5740 - Runner - DEBUG - Using device 0 of [0]
5740 - Runner - DEBUG - Building optimizer and lr scheduler
5740 - Runner - INFO - Found 308 encoder weight params
5740 - Runner - INFO - Found 193 encoder bias params
5740 - Runner - INFO - Found 19 decoder weight params
5740 - Runner - INFO - Found 12 decoder bias params
5740 - Runner - INFO - Loading checkpoint: C:\Users\User\Documents\sdd\saved\sever-FPN-efficientnet-b5-BCEDiceLoss-RAdam\1008_124332\checkpoints\checkpoint-epoch179.pth
5740 - Runner - INFO - Checkpoint "C:\Users\User\Documents\sdd\saved\sever-FPN-efficientnet-b5-BCEDiceLoss-RAdam\1008_124332\checkpoints\checkpoint-epoch179.pth" loaded
5740 - Runner - DEBUG - Getting augmentations
5740 - Runner - DEBUG - Getting data_loader instance
5740 - SamplerFactory - INFO - Creating type...
5740 - SamplerFactory - INFO - Sample population absolute class sizes: [5332 1801]
5740 - SamplerFactory - INFO - Sample population relative class sizes: [0.74751157 0.25248843]
5740 - SamplerFactory - INFO - Target batch class distribution [0.7846383 0.2153617] using alpha=-0.15
5740 - SamplerFactory - INFO - Rounded batch class distribution [0.75 0.25]
5740 - SamplerFactory - INFO - Expecting [6 2] samples of each class per batch, over 891 batches of size 8
5740 - SamplerFactory - INFO - Sampling rates: [1.00262566 0.98945031]
5740 - Runner - DEBUG - Getting loss and metric function handles
5740 - Runner - DEBUG - Initialising trainer
5740 - Trainer - INFO - Freezing encoder weights
5740 - Trainer - INFO - Starting training...
5740 - Trainer - INFO - Unfreezing encoder weights
C:\Users\User\Anaconda3\envs\steel\lib\site-packages\torch\nn\functional.py:1350: UserWarning: nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.
warnings.warn("nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.")
5740 - Trainer - DEBUG - Train Epoch: 0 [0/7128 (0%)] Loss: 0.049221
5740 - Trainer - DEBUG - Train Epoch: 0 [128/7128 (2%)] Loss: 0.051688

I have been experiencing problems while following the training procedures. Please suggest me how to solve the following error.

(sever) D:\RCNN\ssdd>sever train -c experiments/fpn-b5.yml
4608 - Runner - INFO - Using random seed: 447676
4608 - Runner - DEBUG - Building model architecture
Load result: None
4608 - Runner - DEBUG - Using device 0 of [0]
4608 - Runner - DEBUG - Building optimizer and lr scheduler
4608 - Runner - INFO - Found 308 encoder weight params
4608 - Runner - INFO - Found 193 encoder bias params
4608 - Runner - INFO - Found 19 decoder weight params
4608 - Runner - INFO - Found 12 decoder bias params
4608 - Runner - DEBUG - Getting augmentations

demo--------->>>
4608 - Runner - DEBUG - Getting data_loader instance
path: data\raw\severstal-steel-defect-detection
file: train.csv
path: data\raw\severstal-steel-defect-detection
file: pseudo.csv
4608 - SamplerFactory - INFO - Creating type...
4608 - SamplerFactory - INFO - Sample population absolute class sizes: [5332 1801]
4608 - SamplerFactory - INFO - Sample population relative class sizes: [0.74751157 0.25248843]
4608 - SamplerFactory - INFO - Target batch class distribution [0.7846383 0.2153617] using alpha=-0.15
4608 - SamplerFactory - INFO - Rounded batch class distribution [0.77777778 0.22222222]
4608 - SamplerFactory - INFO - Expecting [14 4] samples of each class per batch, over 396 batches of size 18
4608 - SamplerFactory - INFO - Sampling rates: [1.03975994 0.87951138]

Ademo--------->>>
4608 - Runner - DEBUG - Getting loss and metric function handles
4608 - Runner - DEBUG - Initialising trainer
4608 - Trainer - INFO - Freezing encoder weights
4608 - Trainer - INFO - Starting training...
4608 - Trainer - INFO - Unfreezing encoder weights
Traceback (most recent call last):
File "C:\Users\USER\anaconda3\envs\sever\Scripts\sever-script.py", line 33, in
sys.exit(load_entry_point('sever', 'console_scripts', 'sever')())
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\click\core.py", line 829, in call
return self.main(*args, **kwargs)
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\click\core.py", line 782, in main
rv = self.invoke(ctx)
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\click\core.py", line 1259, in invoke
return _process_result(sub_ctx.command.invoke(sub_ctx))
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\click\core.py", line 1066, in invoke
return ctx.invoke(self.callback, **ctx.params)
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\click\core.py", line 610, in invoke
return callback(*args, **kwargs)
File "d:\rcnn\ssdd\sever\cli.py", line 37, in train
Runner(config).train(resume)
File "d:\rcnn\ssdd\sever\main.py", line 77, in train
trainer.train()
File "d:\rcnn\ssdd\sever\base\base_trainer.py", line 59, in train
result = self._train_epoch(epoch)
File "d:\rcnn\ssdd\sever\trainer\trainer.py", line 67, in _train_epoch
for batch_idx, (data, target) in enumerate(self.data_loader):
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\torch\utils\data\dataloader.py", line 819, in next
return self._process_data(data)
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\torch\utils\data\dataloader.py", line 846, in _process_data
data.reraise()
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\torch_utils.py", line 369, in reraise
raise self.exc_type(msg)
AttributeError: Caught AttributeError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\torch\utils\data_utils\worker.py", line 178, in _worker_loop
data = fetcher.fetch(index)
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\torch\utils\data_utils\fetch.py", line 44, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\torch\utils\data_utils\fetch.py", line 44, in
data = [self.dataset[idx] for idx in possibly_batched_index]
File "d:\rcnn\ssdd\sever\data_loader\datasets.py", line 56, in getitem
augmented = self.transforms(image=img, mask=mask)
File "d:\rcnn\ssdd\sever\data_loader\augmentation.py", line 61, in call
return self.transform(*args, **kwargs)
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\albumentations\core\composition.py", line 176, in call
data = t(force_apply=force_apply, **data)
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\albumentations\core\composition.py", line 223, in call
data = t(force_apply=True, **data)
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\albumentations\core\transforms_interface.py", line 87, in call
return self.apply_with_params(params, **kwargs)
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\albumentations\core\transforms_interface.py", line 94, in apply_with_params
params = self.update_params(params, **kwargs)
File "C:\Users\USER\anaconda3\envs\sever\lib\site-packages\albumentations\core\transforms_interface.py", line 142, in update_params
params.update({"cols": kwargs["image"].shape[1], "rows": kwargs["image"].shape[0]})
AttributeError: 'NoneType' object has no attribute 'shape'