Code Monkey home page Code Monkey logo

classification-pytorch's Issues

迭代到50之后就停止了

代码在训练到50个epoch之后就停止了
Uploading 图片.png…
报错是
`Total Loss: 0.733 || Val Loss: 0.695
Epoch 51/100: 0%| | 0/18 [00:00<?, ?it/s<class 'dict'>]Start Train

Process finished with exit code -1073741819 (0xC0000005)`
请问这是什么原因呢?

输入图像对于所有类别的属于可能性的问题

比如说输入img/cat.jpg的时候,预测结果是Class: cat Probability: 0.998,希望可以得到的预测结果是Class: cat Probability: 0.998,dog Probability: 0.002。

#---------------------------------------------------#

获得所属种类

#---------------------------------------------------#
class_name = self.class_names[np.argmax(preds)]
probability = np.max(preds)

我觉得应该是需要修改这块的代码,但是不知道该如何下手...初学者还请多多包涵~
如果能回复我,将不胜感激~~~支持大佬!!!

关于resnet50在不同torchvision下的不同权重问题

https://download.pytorch.org/models/resnet50-19c8e357.pt
https://download.pytorch.org/models/resnet50-0676ba61.pth
我想问下这俩权重为啥不一样,我找到torchvision下的resnet.py文件。发现他们定义的网络结构一样的。为啥加载模型会不一样呢。
要是加载不对应的权重会报错:
RuntimeError: Error(s) in loading state_dict for ResNet:
size mismatch for layer1.0.downsample.1.weight: copying a param with shape torch.Size([256, 64, 1, 1]) from checkpoint, the shape in current model is torch.Size([256]).
size mismatch for layer2.0.downsample.1.weight: copying a param with shape torch.Size([512, 256, 1, 1]) from checkpoint, the shape in current model is torch.Size([512]).
size mismatch for layer3.0.downsample.1.weight: copying a param with shape torch.Size([1024, 512, 1, 1]) from checkpoint, the shape in current model is torch.Size([1024]).
size mismatch for layer4.0.downsample.1.weight: copying a param with shape torch.Size([2048, 1024, 1, 1]) from checkpoint, the shape in current model is torch.Size([2048]).
求大佬指导。

关于DataGenerator的问题

您好,我想问一下,这下面的两个参数True和False是代表DataGenerator类中的ramdom的值吗?
也就是autoagument一直是默认值True
9372940FB7D066844E577AEE62524E11
D0C57D01BEF10340A466E3BC5A83BDF3

数据预处理方式

代码中的数据预处理是resie到 [224,224],之后CenterCrop [224,244],CenterCrop前后图像大小一样是不是意味着CenterCrop不起作用?我看很多其他仓库的代码是训练时候直接Randcrop到 [224,224],而在预测或评估时先resize到 [256,256] 再CenterCrop到 [224,224],想请问一下这个问题。
另外,仓库中的归一化方式是除以127.5,之后-1,把所有像素点归纳到 [-1,1]之间,而其他仓库是除以255,之后用imagenet1K数据集的mean和std作normalize,官方论文中是用哪种方式呢?

运行eval.py出现错误:ValueError: Axis limits cannot be NaN or Inf

不知道怎么解决,不知道哪里出了问题。具体的提示是:
RuntimeWarning: invalid value encountered in true_divide
F1 = (2 * Recall * Precision) / (Recall + Precision)
Save Recall out to metrics_out\Recall.png
Save Precision out to metrics_out\Precision.png
Traceback (most recent call last):
File "F:\ImgClassification\eval.py", line 58, in
top1, top5, Recall, Precision, F1= evaluteTop1_5(classfication, lines, metrics_out_path)

Vit预训练权重

出错信息:

There is no pretrained model for vit_b_16
Traceback (most recent call last):
  File "train.py", line 225, in <module>
    model = get_model_from_name[backbone](input_shape = input_shape, num_classes = num_classes, pretrained = pretrained)
  File "/data/xly/hhs/classification-pytorch-1/nets/vision_transformer.py", line 223, in vit_b_16
    model.load_state_dict(torch.load("/data/xly/hhs/classification-pytorch-1/model_data/vit-patch_16.pth"))
  File "/root/anaconda3/envs/torch38/lib/python3.8/site-packages/torch/serialization.py", line 595, in load
    return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
  File "/root/anaconda3/envs/torch38/lib/python3.8/site-packages/torch/serialization.py", line 749, in _legacy_load
    return legacy_load(f)
  File "/root/anaconda3/envs/torch38/lib/python3.8/site-packages/torch/serialization.py", line 674, in legacy_load
    tar.extract('storages', path=tmpdir)
  File "/root/anaconda3/envs/torch38/lib/python3.8/tarfile.py", line 2272, in extract
    tarinfo = self._get_extract_tarinfo(member, filter_function, path)
  File "/root/anaconda3/envs/torch38/lib/python3.8/tarfile.py", line 2279, in _get_extract_tarinfo
    tarinfo = self.getmember(member)
  File "/root/anaconda3/envs/torch38/lib/python3.8/tarfile.py", line 1962, in getmember
    raise KeyError("filename %r not found" % name)
KeyError: "filename 'storages' not found"

权重文件:
image
环境:
Ubuntu 20.04
其余配置和项目中的requirements一样
image
就是权重文件损坏了吗?

模型参数大小的问题

您好,训练出来的ResNet50 的pth大小为92161kb;
但是作者原文中写的25.6M,这是什么原因呀?
image

大佬,我在Ubuntu上跑训练的时候报这个错误一直找不到解决方法能帮我看看吗

Traceback (most recent call last): File "train.py", line 452, in
fit_one_epoch(model_train, model, loss_history, optimizer, epoch, epoch_step, epoch_step_val, gen, gen_val, UnFreeze_Epoch, Cuda, fp16, save_period, save_dir, local_rank)
File "/mnt/class_master/utils/utils_fit.py", line 116, in fit_one_epoch if (math.floor(epoch) + 1) % save_period == 0 or math.floor(epoch)+1 == Epoch:
TypeError: unsupported operand type(s) for %: 'int' and 'str'

BUS ERROR

你好,在服务器上运行代码中出现bus error,请问是什么错误
image

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.