-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathdraw_pr_fm.py
201 lines (175 loc) · 7.54 KB
/
draw_pr_fm.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
from matplotlib.ticker import MultipleLocator, FormatStrFormatter
from typing import List
import matplotlib.pyplot as plt
import numpy as np
plt.rcParams['font.sans-serif'] = ['Microsoft YaHei'] # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False # 用来正常显示负号
def draw_pr_fm(npy_path_list: List, model_name_list: List, save_path=''):
'''
将各个模型得到recall-precision、fppi-mr曲线聚合绘制到两个子图中
:param npy_path_list: 各个模型测试过程中产生的npy文件路径列表
:param model_name_list: 各个模型对应的名字
:param save_path: 若非空则指向当前图像的保存路径
:return:
'''
assert len(npy_path_list) == len(model_name_list)
pr_model_names = model_name_list.copy()
fm_model_names = model_name_list.copy()
plt.figure(figsize=(13, 5), dpi=300)
plt.subplots_adjust(left=0.05, right=0.98, bottom=0.15, top=0.98, wspace=0.15, hspace=0)
for i, (npy_path, _) in enumerate(zip(npy_path_list, model_name_list)):
# 从各个模型对应的npy文件中取出recall、precision、fppi和mr数据
ld: dict = np.load(npy_path, allow_pickle=True).item()
rec = ld['recall']
prec = ld['precision']
fppi = ld['fppi']
lamr = ld['lamr']
mr = ld['mr']
ap = ld['ap']
# 为各个模型的名字中添加AP和LAMR指标
pr_model_names[i] += f" ({(ap * 100):.1f}%)"
fm_model_names[i] += f" ({(lamr * 100):.1f}%)"
# 绘制PR曲线
plt.subplot(1, 2, 1)
plt.plot(rec, prec)
# 绘制FPPI-MR曲线
plt.subplot(1, 2, 2)
plt.plot(fppi, mr)
# 设置子图图例、标题、坐标轴标签、网格等各项参数
plt.subplot(1, 2, 1)
plt.legend(pr_model_names, frameon=True, loc='lower left')
plt.title("(a) P-R 曲线", y=-0.18)
plt.xlabel("Recall")
plt.ylabel("Precision")
plt.grid(ls='--')
plt.subplot(1, 2, 2)
plt.legend(fm_model_names, frameon=True, loc='lower left')
plt.title("(b) FPPI-MR 曲线", y=-0.18)
plt.xlabel("False Positives Per Image")
plt.ylabel("Miss Rate")
# 将FPPI-MR图绘制成log-log坐标格式的图像
plt.xscale('log')
plt.yscale('log')
ax = plt.gca()
# 使FPPI-MR图的坐标轴以0.1为主刻度距离，并限制y轴刻度范围
ax.yaxis.set_major_locator(MultipleLocator(0.1))
ax.yaxis.set_major_formatter(FormatStrFormatter('%.1f'))
plt.ylim(0.1, 1)
plt.grid(ls='--')
# 保存或者展示P-R/FPPI-MR生成图
if save_path == '':
plt.show()
else:
plt.savefig(save_path)
def draw_pr(npy_path_list: List, model_name_list: List, save_path=''):
'''
绘制各个模型得到recall-precision曲线
:param npy_path_list: 各个模型测试过程中产生的npy文件路径列表
:param model_name_list: 各个模型对应的名字
:param save_path: 若非空则指向当前图像的保存路径
:return:
'''
assert len(npy_path_list) == len(model_name_list)
pr_model_names = model_name_list.copy()
plt.figure(figsize=(7, 6), dpi=300)
plt.subplots_adjust(left=0.1, right=0.95, bottom=0.1, top=0.95, wspace=0.15, hspace=0)
for i, (npy_path, _) in enumerate(zip(npy_path_list, model_name_list)):
# 从各个模型对应的npy文件中取出recall、precision、fppi和mr数据
ld: dict = np.load(npy_path, allow_pickle=True).item()
rec = ld['recall']
prec = ld['precision']
ap = ld['ap']
pr_model_names[i] += f" ({(ap * 100):.1f}%)"
plt.plot(rec, prec)
plt.legend(pr_model_names, frameon=True, loc='lower left')
# plt.title("P-R 曲线", y=-0.18)
plt.xlabel("Recall")
plt.ylabel("Precision")
plt.grid(ls='--')
if save_path == "":
plt.show()
else:
plt.savefig(save_path)
def draw_fppi_mr(npy_path_list: List, model_name_list: List, save_path=''):
'''
绘制各个模型得到的fppi-mr曲线
:param npy_path_list: 各个模型测试过程中产生的npy文件路径列表
:param model_name_list: 各个模型对应的名字
:param save_path: 若非空则指向当前图像的保存路径
:return:
'''
assert len(npy_path_list) == len(model_name_list)
fm_model_names = model_name_list.copy()
plt.figure(figsize=(7, 6), dpi=300)
plt.subplots_adjust(left=0.1, right=0.95, bottom=0.1, top=0.95, wspace=0.15, hspace=0)
for i, (npy_path, _) in enumerate(zip(npy_path_list, model_name_list)):
# 从各个模型对应的npy文件中取出recall、precision、fppi和mr数据
ld: dict = np.load(npy_path, allow_pickle=True).item()
fppi = ld['fppi']
lamr = ld['lamr']
mr = ld['mr']
fm_model_names[i] += f" ({(lamr * 100):.1f}%)"
# 绘制FPPI-MR曲线
plt.plot(fppi, mr)
plt.legend(fm_model_names, frameon=True, loc='lower left')
# plt.title("(b) FPPI-MR 曲线", y=-0.18)
plt.xlabel("False Positives Per Image")
plt.ylabel("Miss Rate")
# 将FPPI-MR图绘制成log-log坐标格式的图像
plt.xscale('log')
plt.yscale('log')
ax = plt.gca()
# 使FPPI-MR图的坐标轴以0.1为主刻度距离，并限制y轴刻度范围
ax.yaxis.set_major_locator(MultipleLocator(0.1))
ax.yaxis.set_major_formatter(FormatStrFormatter('%.1f'))
plt.ylim(0.1, 1)
plt.grid(ls='--')
if save_path == '':
plt.show()
else:
plt.savefig(save_path)
if __name__ == '__main__':
yolov3_npy_paths = [
"results/Visible-YOLOv3-Normal102/rec-prec.fppi-mr.npy",
"results/Double-YOLOv3-Add-SL102/rec-prec.fppi-mr.npy",
"results/Double-YOLOv3-Concat-SE102/rec-prec.fppi-mr.npy",
# "results/Double-YOLOv3-Fshare-Global-Add-SL102/rec-prec.fppi-mr.npy",
# "results/Double-YOLOv3-Fshare-Global-Concat-SE102/rec-prec.fppi-mr.npy",
# "results/Double-YOLOv3-Fshare-Global-Concat-SE3-102/rec-prec.fppi-mr.npy",
]
yolov3_model_names = [
"Visible-YOLOv3",
"Double-YOLOv3-ASL",
"Double-YOLOv3-CSE",
# "Double-YOLOv3-FSHASL",
# "Double-YOLOv3-FSHCSE5",
# "Double-YOLOv3_FSHCSE"
]
draw_pr(yolov3_npy_paths, yolov3_model_names, save_path="")
draw_fppi_mr(yolov3_npy_paths, yolov3_model_names, save_path="")
yolov4_npy_paths = [
"results/Visible-YOLOv4-Normal102/rec-prec.fppi-mr.pny.npy",
# "results/Visible-YOLOv4-MNv2-102/rec-prec.fppi-mr.npy",
# "results/Visible-YOLOv4-MNv3-102/rec-prec.fppi-mr.npy",
"results/Double-YOLOv4-Concat-SE102/rec-prec.fppi-mr.npy",
"results/Double-YOLOv4-Add-SL102/rec-prec.fppi-mr.npy",
# "results/Double-YOLOv4-Fshare-Global-Concat-SE3v-102/rec-prec.fppi-mr.npy",
# "results/Double-YOLOv4-MNv2-Fshare-Global-CSE3-102/rec-prec.fppi-mr.npy",
# "results/Double-YOLOv4-MNv3-Fshare-Global-CSE3-102/rec-prec.fppi-mr.npy"
]
yolov4_model_names = [
"Visible-YOLOv4",
# "V-YOLOv4-MN2",
# "V-YOLOv4-MN3",
"Double-YOLOv4-ASL",
"Double-YOLOv4-CSE",
# "Double-YOLOv4-FSHCSE",
# "D-YOLOv4-MNv2-FSHCSE",
# "D-YOLOv4-MNv3-FSHCSE"
]
draw_pr(yolov4_npy_paths, yolov4_model_names, save_path="")
draw_fppi_mr(yolov4_npy_paths, yolov4_model_names, save_path="")
draw_pr(yolov3_npy_paths + yolov4_npy_paths, yolov3_model_names + yolov4_model_names,
save_path="docs/yolov3-4.pr6.png")
draw_fppi_mr(yolov3_npy_paths + yolov4_npy_paths, yolov3_model_names + yolov4_model_names,
save_path="docs/yolov3-4.fm6.png")