前言:【从零开始学习YOLOv3】系列越写越多,本来安排的内容比较少,但是在阅读代码的过程中慢慢发掘了一些新的亮点,所以不断加入到这个系列中。之前都在读YOLOv3中的代码,已经学习了cfg文件、模型构建等内容。本文在之前的基础上,对模型的代码进行修改,将之前Attention系列中的SE模块和CBAM模块集成到YOLOv3中。
1. 规定格式
正如[convolutional]
,[maxpool]
,[net]
,[route]
等层在cfg中的定义一样,我们再添加全新的模块的时候,要规定一下cfg的格式。做出以下规定:
在SE模块(具体讲解见: 【cv中的Attention机制】最简单最易实现的SE模块)中,有一个参数为reduction
,这个参数默认是16,所以在这个模块中的详细参数我们按照以下内容进行设置:
[se]
reduction=16
在CBAM模块(具体讲解见: 【CV中的Attention机制】ECCV 2018 Convolutional Block Attention Module)中,空间注意力机制和通道注意力机制中一共存在两个参数:ratio
和kernel_size
, 所以这样规定CBAM在cfg文件中的格式:
[cbam]
ratio=16
kernelsize=7
2. 修改解析部分
由于我们添加的这些参数都是自定义的,所以需要修改解析cfg文件的函数,之前讲过,需要修改parse_config.py
中的部分内容:
def parse_model_cfg(path):
# path参数为: cfg/yolov3-tiny.cfg
if not path.endswith('.cfg'):
path += '.cfg'
if not os.path.exists(path) and \
os.path.exists('cfg' + os.sep + path):
path = 'cfg' + os.sep + path
with open(path, 'r') as f:
lines = f.read().split('\n')
# 去除以#开头的,属于注释部分的内容
lines = [x for x in lines if x and not x.startswith('#')]
lines = [x.rstrip().lstrip() for x in lines]
mdefs = [] # 模块的定义
for line in lines:
if line.startswith('['): # 标志着一个模块的开始
'''
eg:
[shortcut]
from=-3
activation=linear
'''
mdefs.append({})
mdefs[-1]['type'] = line[1:-1].rstrip()
if mdefs[-1]['type'] == 'convolutional':
mdefs[-1]['batch_normalize'] = 0
else:
key, val = line.split("=")
key = key.rstrip()
if 'anchors' in key:
mdefs[-1][key] = np.array([float(x) for x in val.split(',')]).reshape((-1, 2))
else:
mdefs[-1][key] = val.strip()
# Check all fields are supported
supported = ['type', 'batch_normalize', 'filters', 'size',\
'stride', 'pad', 'activation', 'layers', \
'groups','from', 'mask', 'anchors', \
'classes', 'num', 'jitter', 'ignore_thresh',\
'truth_thresh', 'random',\
'stride_x', 'stride_y']
f = [] # fields
for x in mdefs[1:]:
[f.append(k) for k in x if k not in f]
u = [x for x in f if x not in supported] # unsupported fields
assert not any(u), "Unsupported fields %s in %s. See https://github.com/ultralytics/yolov3/issues/631" % (u, path)
return mdefs
以上内容中,需要改的是supported中的字段,将我们的内容添加进去:
supported = ['type', 'batch_normalize', 'filters', 'size',\
'stride', 'pad', 'activation', 'layers', \
'groups','from', 'mask', 'anchors', \
'classes', 'num', 'jitter', 'ignore_thresh',\
'truth_thresh', 'random',\
'stride_x', 'stride_y',\
'ratio', 'reduction', 'kernelsize']
3. 实现SE和CBAM
具体原理还请见【cv中的Attention机制】最简单最易实现的SE模块和【CV中的Attention机制】ECCV 2018 Convolutional Block Attention Module这两篇文章,下边直接使用以上两篇文章中的代码:
SE
class SELayer(nn.Module):
def __init__(self, channel, reduction=16):
super(SELayer, self).__init__()
self.avg_pool = nn.AdaptiveAvgPool2d(1)
self.fc = nn.Sequential(
nn.Linear(channel, channel // reduction, bias=False),
nn.ReLU(inplace=True),
nn.Linear(channel // reduction, channel, bias=False),
nn.Sigmoid()
)
def forward(self, x):
b, c, _, _ = x.size()
y = self.avg_pool(x).view(b, c)
y = self.fc(y).view(b, c, 1, 1)
return x * y.expand_as(x)
CBAM
class SpatialAttention(nn.Module):
def __init__(self, kernel_size=7):
super(SpatialAttention, self).__init__()
assert kernel_size in (3,7), "kernel size must be 3 or 7"
padding = 3if kernel_size == 7else1
self.conv = nn.Conv2d(2,1,kernel_size, padding=padding, bias=False)
self.sigmoid = nn.Sigmoid()
def forward(self, x):
avgout = torch.mean(x, dim=1, keepdim=True)
maxout, _ = torch.max(x, dim=1, keepdim=True)
x = torch.cat([avgout, maxout], dim=1)
x = self.conv(x)
return self.sigmoid(x)
class ChannelAttention(nn.Module):
def __init__(self, in_planes, rotio=16):
super(ChannelAttention, self).__init__()
self.avg_pool = nn.AdaptiveAvgPool2d(1)
self.max_pool = nn.AdaptiveMaxPool2d(1)
self.sharedMLP = nn.Sequential(
nn.Conv2d(in_planes, in_planes // ratio, 1, bias=False), nn.ReLU(),
nn.Conv2d(in_planes // rotio, in_planes, 1, bias=False))
self.sigmoid = nn.Sigmoid()
def forward(self, x):
avgout = self.sharedMLP(self.avg_pool(x))
maxout = self.sharedMLP(self.max_pool(x))
return self.sigmoid(avgout + maxout)
以上就是两个模块的代码,添加到models.py
文件中。
4. 设计cfg文件
这里以yolov3-tiny.cfg
为baseline,然后添加注意力机制模块。
CBAM与SE类似,所以以SE为例,添加到backbone之后的部分,进行信息重构(refinement)。
[net]
# Testing
batch=1
subdivisions=1
# Training
# batch=64
# subdivisions=2
width=416
height=416
channels=3
momentum=0.9
decay=0.0005
angle=0
saturation = 1.5
exposure = 1.5
hue=.1
learning_rate=0.001
burn_in=1000
max_batches = 500200
policy=steps
steps=400000,450000
scales=.1,.1
[convolutional]
batch_normalize=1
filters=16
size=3
stride=1
pad=1
activation=leaky
[maxpool]
size=2
stride=2
[convolutional]
batch_normalize=1
filters=32
size=3
stride=1
pad=1
activation=leaky
[maxpool]
size=2
stride=2
[convolutional]
batch_normalize=1
filters=64
size=3
stride=1
pad=1
activation=leaky
[maxpool]
size=2
stride=2
[convolutional]
batch_normalize=1
filters=128
size=3
stride=1
pad=1
activation=leaky
[maxpool]
size=2
stride=2
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[maxpool]
size=2
stride=2
[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky
[maxpool]
size=2
stride=1
[convolutional]
batch_normalize=1
filters=1024
size=3
stride=1
pad=1
activation=leaky
[se]
reduction=16
# 在backbone结束的地方添加se模块
#####backbone######
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky
[convolutional]
size=1
stride=1
pad=1
filters=18
activation=linear
[yolo]
mask = 3,4,5
anchors = 10,14, 23,27, 37,58, 81,82, 135,169, 344,319
classes=1
num=6
jitter=.3
ignore_thresh = .7
truth_thresh = 1
random=1
[route]
layers = -4
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[upsample]
stride=2
[route]
layers = -1, 8
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[convolutional]
size=1
stride=1
pad=1
filters=18
activation=linear
[yolo]
mask = 0,1,2
anchors = 10,14, 23,27, 37,58, 81,82, 135,169, 344,319
classes=1
num=6
jitter=.3
ignore_thresh = .7
truth_thresh = 1
random=1
5. 模型构建
以上都是准备工作,以SE为例,我们修改model.py
文件中的模型加载部分,并修改forward函数部分的代码,让其正常发挥作用:
在model.py
中的create_modules
函数中进行添加:
elif mdef['type'] == 'se':
modules.add_module(
'se_module',
SELayer(output_filters[-1], reduction=int(mdef['reduction'])))
然后修改Darknet中的forward部分的函数:
def forward(self, x, var=None):
img_size = x.shape[-2:]
layer_outputs = []
output = []
for i, (mdef,
module) in enumerate(zip(self.module_defs, self.module_list)):
mtype = mdef['type']
if mtype in ['convolutional', 'upsample', 'maxpool']:
x = module(x)
elif mtype == 'route':
layers = [int(x) for x in mdef['layers'].split(',')]
if len(layers) == 1:
x = layer_outputs[layers[0]]
else:
try:
x = torch.cat([layer_outputs[i] for i in layers], 1)
except: # apply stride 2 for darknet reorg layer
layer_outputs[layers[1]] = F.interpolate(
layer_outputs[layers[1]], scale_factor=[0.5, 0.5])
x = torch.cat([layer_outputs[i] for i in layers], 1)
elif mtype == 'shortcut':
x = x + layer_outputs[int(mdef['from'])]
elif mtype == 'yolo':
output.append(module(x, img_size))
layer_outputs.append(x if i in self.routs else [])
在forward中加入SE模块,其实很简单。SE模块与卷积层,上采样,最大池化层地位是一样的,不需要更多操作,只需要将以上部分代码进行修改:
for i, (mdef,
module) in enumerate(zip(self.module_defs, self.module_list)):
mtype = mdef['type']
if mtype in ['convolutional', 'upsample', 'maxpool', 'se']:
x = module(x)
CBAM的整体过程类似,可以自己尝试一下,顺便熟悉一下YOLOv3的整体流程。
后记:本文的内容很简单,只是添加了注意力模块,很容易实现。不过具体注意力机制的位置、放多少个模块等都需要做实验来验证。注意力机制并不是万金油,需要多调参,多尝试才能得到满意的结果。欢迎大家联系我加入群聊,反馈在各自数据集上的效果。
ps: 最近大家注意身体,出门戴口罩。
来源:oschina
链接:https://my.oschina.net/u/4357035/blog/3313596