【深度学习】Yolo记录

yolov4记录

bboxes_pred = model(images) #bboxes_pred为列表，长度为3，三个尺度的预测
'''例如sub batch_size为4时 类别为5时
bboxes_pred[0].shape -> torch.Size([4, 30, 76, 76])
bboxes_pred[1].shape -> torch.Size([4, 30, 38, 38])
bboxes_pred[2].shape -> torch.Size([4, 30, 19, 19])'''

loss, loss_xy, loss_wh, loss_obj, loss_cls, loss_l2 = criterion(bboxes_pred, bboxes)
'''
bboxes.shape -> torch.Size([4, 60, 5])
'''

关于模型的输出：
三个尺度预测，每个尺度的stride分别为[8, 16, 32]，输入图片大小为608，因此特征维度分别为[76, 38, 19]，每个尺度的格子预测三个anchor
例如尺度1，预测76*76个格子，每个格子对3个anchor进行回归

yolov5模型记录

以下均为yolov5s的有关记录

关于标签

1.train.txt test.txt训练测试文件 列举出所有参与训练的图片及其路径 每一行一个（修改的为txt + jpg）
2.每一张图片对应一个txt标签文件，每一行代表一个目标
格式为目标序号 x y w h，目标序号从零开始，不代表背景，xy为中心点， xywh均为相对值（0-1）例如：

0 0.9250814332247557 0.6224489795918368 0.013029315960912053 0.013605442176870748
2 0.3127035830618892 0.4736394557823129 0.013029315960912053 0.015306122448979591
2 0.08224755700325732 0.641156462585034 0.014657980456026058 0.013605442176870748

模型网络打印

                 from  n    params  module                                  arguments                     
  0                -1  1      3520  models.common.Focus                     [3, 32, 3]                    
  1                -1  1     18560  models.common.Conv                      [32, 64, 3, 2]                
  2                -1  1     19904  models.common.BottleneckCSP             [64, 64, 1]                   
  3                -1  1     73984  models.common.Conv                      [64, 128, 3, 2]               
  4                -1  1    161152  models.common.BottleneckCSP             [128, 128, 3]                 
  5                -1  1    295424  models.common.Conv                      [128, 256, 3, 2]              
  6                -1  1    641792  models.common.BottleneckCSP             [256, 256, 3]                 
  7                -1  1   1180672  models.common.Conv                      [256, 512, 3, 2]              
  8                -1  1    656896  models.common.SPP                       [512, 512, [5, 9, 13]]        
  9                -1  1   1248768  models.common.BottleneckCSP             [512, 512, 1, False]          
 10                -1  1    131584  models.common.Conv                      [512, 256, 1, 1]              
 11                -1  1         0  torch.nn.modules.upsampling.Upsample    [None, 2, 'nearest']          
 12           [-1, 6]  1         0  models.common.Concat                    [1]                           
 13                -1  1    378624  models.common.BottleneckCSP             [512, 256, 1, False]          
 14                -1  1     33024  models.common.Conv                      [256, 128, 1, 1]              
 15                -1  1         0  torch.nn.modules.upsampling.Upsample    [None, 2, 'nearest']          
 16           [-1, 4]  1         0  models.common.Concat                    [1]                           
 17                -1  1     95104  models.common.BottleneckCSP             [256, 128, 1, False]          
 18                -1  1    147712  models.common.Conv                      [128, 128, 3, 2]              
 19          [-1, 14]  1         0  models.common.Concat                    [1]                           
 20                -1  1    313088  models.common.BottleneckCSP             [256, 256, 1, False]          
 21                -1  1    590336  models.common.Conv                      [256, 256, 3, 2]              
 22          [-1, 10]  1         0  models.common.Concat                    [1]                           
 23                -1  1   1248768  models.common.BottleneckCSP             [512, 512, 1, False]          
 24      [17, 20, 23]  1    229245  Detect                                  [80, [[10, 13, 16, 30, 33, 23], [30, 61, 62, 45, 59, 119], [116, 90, 156, 198, 373, 326]], [128, 256, 512]]

训练时的输入输出

model.train()
img = torch.rand(8 if torch.cuda.is_available() else 1, 3, 640, 640).to(device)
y = model(img, profile=False)
print(type(y))	# <class 'list'>
print(len(y))	# 3尺度输出
print(type(y[0]))	# <class 'torch.Tensor'>
print(y[0].size())	# torch.Size([8, 3, 80, 80, 85])
print(y[1].size())	# torch.Size([8, 3, 40, 40, 85])
print(y[2].size())	# torch.Size([8, 3, 20, 20, 85])

评估时的输入输出

评估时输出变为列表，多加了一个输出处于0位，原训练输出放在了列表的1位

model.eval()
print(type(y))	# <class 'tuple'>
print(len(y))	# 2
print(type(y[0]))	# <class 'torch.Tensor'>
print(type(y[1]))	# <class 'list'>
print(y[0].size())	# torch.Size([8, 25200, 85])
print(len(y[1]))	# 3
print(y[1][0].size())	# torch.Size([8, 3, 80, 80, 85])
print(y[1][1].size())	# torch.Size([8, 3, 40, 40, 85])
print(y[1][2].size())	# torch.Size([8, 3, 20, 20, 85])

# 输入1×3×2560×2560
<class 'tuple'>
2
<class 'torch.Tensor'>
<class 'list'>
torch.Size([1, 403200, 85])
3
torch.Size([1, 3, 320, 320, 85])
torch.Size([1, 3, 160, 160, 85])
torch.Size([1, 3, 80, 80, 85])

anchors有关

s = 128  # 2x min stride
m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))])  # forward
print('m.stride', m.stride)     # tensor([ 8., 16., 32.])
print('for', [x.shape for x in self.forward(torch.zeros(1, ch, s, s))]) # [torch.Size([1, 3, 16, 16, 85]), torch.Size([1, 3, 8, 8, 85]), torch.Size([1, 3, 4, 4, 85])]
            
- [10,13, 16,30, 33,23]  # P3/8
- [30,61, 62,45, 59,119]  # P4/16
- [116,90, 156,198, 373,326]  # P5/32
# anchors 会在模型创建时 对anchors会再做一次处理
m.anchors /= m.stride.view(-1, 1, 1)
print('m.anchors', m.anchors)
tensor([[[ 1.25000,  1.62500],
         [ 2.00000,  3.75000],
         [ 4.12500,  2.87500]],

        [[ 1.87500,  3.81250],
         [ 3.87500,  2.81250],
         [ 3.68750,  7.43750]],

        [[ 3.62500,  2.81250],
         [ 4.87500,  6.18750],
         [11.65625, 10.18750]]])