Skip to content

issues Search Results · repo:Res2Net/Res2Net-PretrainedModels language:Python

Filter by

74 results
 (65 ms)

74 results

inRes2Net/Res2Net-PretrainedModels (press backspace or delete to remove)

Res2Net/Res2Net-PretrainedModels
Code Copyright Issues

Hi, I saw on a paper that he used the res2net_v1b code you posted, and I m not sure if this paper was published by you, so I wanted to ask if I can use the res2net_v1b code you posted for my paper experiments? ...
  • Summerx026
  • 2
  • Opened 
    on Jan 30, 2024
  • #77

Res2Net/Res2Net-PretrainedModels
代码中的残差结构和图像中不一致

代码： spx = torch.split(out, self.width, 1) for i in range(self.nums): if i==0 or self.stype== stage : sp = spx[i] else: sp = sp + spx[i] sp = self.convsi sp = self.relu(self.bnsi) if i==0: out = sp else: ...
  • VagoLapis
  • 1
  • Opened 
    on Jan 11, 2024
  • #76

Res2Net/Res2Net-PretrainedModels
self-supervised training on Imagenet

Is it possible to provide the self-supervised pre-trained model? Specifically, the model is trained on Imagenet without using the ground truth label. The model is trained in an self-supervised manner, ...
  • ForawardStar
  • Opened 
    on Oct 11, 2023
  • #75

Res2Net/Res2Net-PretrainedModels
Problem of Replacement of "BatchNorm" to "GroupNorm"

Hi, Dr. Gao, I have a question about model modification. Due to my work setting, I could only set batch_size=1, so I am going to modify the BatchNorm to GroupNorm. Based on the original code, I just changed ...
  • CinKKKyo
  • Opened 
    on Jun 7, 2023
  • #74

Res2Net/Res2Net-PretrainedModels
对于每一个layer的第一个block，计算方式不是层级的，和论文中描述有差别

第一个block计算方式
  • easy2star
  • 3
  • Opened 
    on May 6, 2023
  • #73

Res2Net/Res2Net-PretrainedModels
About SSLD pretrained model

Hi, I am the author of SSLD. I noticed that the weight of SSLD has been added to your repo, but here it should be Res2Net-D instead of ResNet-B. I hope you can confirm it at your convenience.
  • cuicheng01
  • 3
  • Opened 
    on Dec 8, 2022
  • #72

Res2Net/Res2Net-PretrainedModels
res2net50-v1b的训练策略

您好，我现在在复现res2net在imagenet数据集上的分类精度，我参考了论文，res2net50的训练策略如下： scheduler: step_decay lr: 0.1 warmup_epochs: 0 decay_rate: 0.1 decay_epochs: 30 epoch_size: 100 res2net精度可以达到top1: 78.00%，和论文是相符的； 在此基础上添加了cutmix数据增强策略，epoch_size变为300，每75epoch学习率下降为原来的0.1，来训练res2net50-v1b；res2net50-v1b实测精度只有79.80%，与论文相比还差0.5%。因为代码中没有提供训练的参数，可能在某些地方没对齐，想问一下能不能分享一下res2net50-v1b的训练策略或者训练参数。谢谢。 ...
  • zp5070
  • 2
  • Opened 
    on Nov 2, 2022
  • #71

Res2Net/Res2Net-PretrainedModels
26是否可以替换成其他更小数值

看res2net的layer1部分：26*4 = 104 ==conv1x1== 256 想减小通道数，把256 改成 64 请问，可以把26改成13，或者更小吗？会有什么不好的影响吗 比如下面： ① 134 = 52 ==conv1x1== 64 ② 74 = 28 ==conv1x1== 64 谢谢
  • laocan
  • 2
  • Opened 
    on Sep 23, 2022
  • #70

Res2Net/Res2Net-PretrainedModels
why don't have the Res2Net-v1b-200-SSLD model code ?

  • wszhengjx
  • 1
  • Opened 
    on Sep 13, 2022
  • #69

Res2Net/Res2Net-PretrainedModels
张量尺寸不匹配问题

你好请问 for i in range(self.nums): if i==0 or self.stype== stage : sp = spx[i] else: sp = sp + spx[i] //这里因为两个特征的尺寸不一样并不能加在一起啊？一个是64一个是128
  • kirkzZ
  • 1
  • Opened 
    on Aug 8, 2022
  • #68
Issue origami icon

Learn how you can use GitHub Issues to plan and track your work.

Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub Issues
ProTip! 
Restrict your search to the title by using the in:title qualifier.
Issue origami icon

Learn how you can use GitHub Issues to plan and track your work.

Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub Issues
ProTip! 
Restrict your search to the title by using the in:title qualifier.
Issue search results · GitHub