YOLOv5+Swin Transformer

本科生工科生cv改代码

本来做的7，但是7报错一直解决不了，我就试试5

1、先是第一个报错

TypeError: __init__() missing 1 required positional argument: 'c2'

解决：在yolo.py里

1.  
   if m in {
2.  
   Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv,
3.  
   BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x, C3STR}:
4.  
   # add a C3STR at the end of the sentence(aran)
5.  
   c1, c2 = ch[f], args[0]
6.  
   if c2 != no: # if not output
7.  
   c2 = make_divisible(c2 * gw, 8) # change 8 to 9
8.  
    
9.  
   args = [c1, c2, *args[1:]]
10.  
    if m in {BottleneckCSP, C3, C3TR, C3Ghost, C3x, C3STR}:# add a C3STR at the end of the sentence(aran)
11.  
    args.insert(2, n) # number of repeats

2、

1.  
   File "/root/yolov5_master/models/common.py", line 1315, in __init__
2.  
   super().__init__(c1, c2, c2, n, shortcut, g, e)
3.  
   TypeError: __init__() takes from 3 to 7 positional arguments but 8 were given

解决：common里删掉一个c2

1.  
   class C3STR(C3):
2.  
   # C3 module with SwinTransformerBlock()
3.  
   def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):
4.  
   super().__init__(c1, c2, n, shortcut, g, e) # a c2 was deleted by aran
5.  
   c_ = int(c2 * e)
6.  
   num_heads = c_ // 32
7.  
   self.m = SwinTransformerBlock(c_, c_, num_heads, n)

3、

NameError: name 'window_partition' is not defined

解决：应该加上window_partition 和 window_reverse, 加在common，具体位置是在芒果哥加在common的代码前面

1.  
   def window_partition(x, window_size):
2.  
   B, H, W, C = x.shape
3.  
   x = x.view(B, H // window_size, window_size, W // window_size, window_size, C)
4.  
   windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)
5.  
   return windows
6.  
    
7.  
    
8.  
   def window_reverse(windows, window_size, H, W):
9.  
   B = int(windows.shape[0] / (H * W / window_size / window_size))
10.  
    x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1)
11.  
    x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1)
12.  
    return x

4、

NameError: name 'F' is not defined

解决：在common里面

import torch.nn.functional as F

这篇好文章是转载于：学新通技术网