1. 中文
    1. 简介
    2. 实验
    3. 相关工作和讨论
    4. 结论
  2. English
    1. Introduction
    2. Related work
      1. Traffic Sign Classification
      2. Object Detection by CNNs
    3. Benchmark
      1. Data Collection
      2. Data Annotation
      3. Dataset Statistics
    4. Neural Network
      1. Architecture
      2. Training
    5. Results
      1. Detection
      2. Simultaneous detection and classification
    6. Conclusions
  3. 数据分析 Data Analysis
    1. train
    2. test

(CVPR 2016) Traffic-Sign Detection and Classification in the Wild

Paper:

Page:

Code:

我们做了两个贡献:

  1. 我们从 10 万张腾讯街景全景图中创建了一个大型交通标志基准,超越了之前的基准。 它提供包含 30000 个交通标志实例的 100000 张图像。 这些图像涵盖了照度和天气条件的巨大变化。基准测试中的每个交通标志都标注了类别标签、边界框和像素掩码。 我们将此基准称为 Tsinghua-Tencent 100K。

  2. 我们演示了一个强大的端到端卷积神经网络(CNN)如何同时检测和分类交通标志。大多数先前的 CNN 图像处理解决方案针对占据图像的大部分的对象,并且这种网络不能很好地用于仅占据图像的一小部分的目标对象,比如这里的交通标志。实验结果表明了我们网络的稳健性及其对替代方案的优越性。

手机上的澳门永利真的假的We make two contributions:

  1. 手机上的澳门永利真的假的we have created a large traffic-sign benchmark from 100000 Tencent Street View panoramas, going beyond previous benchmarks. It provides 100000 images containing 30000 traffic-sign instances. These images cover large variations in illuminance and weather conditions. Each traffic-sign in the benchmark is annotated with a class label, its bounding box and pixel mask. We call this benchmark Tsinghua-Tencent 100K.

  2. 手机上的澳门永利真的假的we demonstrate how a robust end-to-end convolutional neural network (CNN) can simultaneously detect and classify traffic-signs. Most previous CNN image processing solutions target objects that occupy a large proportion of an image, and such networks do not work well for target objects occupying only a small fraction of an image like the traffic-signs here. Experimental results show the robustness of our network and its superiority to alternatives.

中文

简介

实验

相关工作和讨论

结论

English

Introduction

Traffic Sign Classification

Object Detection by CNNs

Benchmark

Data Collection

Data Annotation

Dataset Statistics

Neural Network

Architecture

Training

Results

Detection

Simultaneous detection and classification

Conclusions

数据分析 Data Analysis

num_classes: 222 (0 background + 221 traffic signs)

1 i1 2 i2 3 i3 4 i4 5 i5 6 i6 7 i7 8 i8 9 i9 10 i10
11 i11 12 i12 13 i13 14 i14 15 i15 16 il50 17 il60 18 il70 19 il80 20 il90
21 il100 22 il110 23 ilx 24 io 25 ip 26 p1 27 p2 28 p3 29 p4 30 p5
31 p6 32 p7 33 p8 34 p9 35 p10 36 p11 37 p12 38 p13 39 p14 40 p15
41 p16 42 p17 43 p18 44 p19 45 p20 46 p21 47 p22 48 p23 49 p24 50 p25
51 p26 52 p27 53 p28 54 p29 55 pa8 56 pa10 57 pa12 58 pa13 59 pa14 60 pax
61 pb 62 pc 63 pd 64 pe 65 pg 66 ph1.5 67 ph2 68 ph2.1 69 ph2.2 70 ph2.4
71 ph2.5 72 ph2.6 73 ph2.8 74 ph2.9 75 ph3 76 ph3.2 77 ph3.3 78 ph3.5 79 ph3.8 80 ph4
81 ph4.2 82 ph4.3 83 ph4.4 84 ph4.5 85 ph4.8 86 ph5 87 ph5.3 88 ph5.5 89 phx 90 pl0
91 pl3 92 pl4 93 pl5 94 pl10 95 pl15 96 pl20 97 pl25 98 pl30 99 pl35 100 pl40
101 pl50 102 pl60 103 pl65 104 pl70 105 pl80 106 pl90 107 pl100 108 pl110 109 pl120 110 plx
111 pm1.5 112 pm2 113 pm2.5 114 pm5 115 pm8 116 pm10 117 pm13 118 pm15 119 pm20 120 pm25
121 pm30 122 pm35 123 pm40 124 pm46 125 pm50 126 pm55 127 pmx 128 pn 129 pn40 130 pne
131 pnl 132 po 133 pr10 134 pr20 135 pr30 136 pr40 137 pr45 138 pr50 139 pr60 140 pr70
141 pr80 142 pr100 143 prx 144 ps 145 pw2 146 pw2.5 147 pw3 148 pw3.2 149 pw3.5 150 pw4
151 pw4.2 152 pw4.5 153 pwx 154 w1 155 w2 156 w3 157 w4 158 w5 159 w6 160 w7
161 w8 162 w9 163 w10 164 w11 165 w12 166 w13 167 w14 168 w15 169 w16 170 w17
171 w18 172 w19 173 w20 174 w21 175 w22 176 w23 177 w24 178 w25 179 w26 180 w27
181 w28 182 w29 183 w30 184 w31 185 w32 186 w33 187 w34 188 w35 189 w36 190 w37
191 w38 192 w39 193 w40 194 w41 195 w42 196 w43 197 w44 198 w45 199 w46 200 w47
201 w48 202 w49 203 w50 204 w51 205 w52 206 w53 207 w54 208 w55 209 w56 210 w57
211 w58 212 w59 213 w60 214 w61 215 w62 216 w63 217 w64 218 w65 219 w66 220 w67
221 wo

train

手机上的澳门永利真的假的train: 7196 samples (train 6105 samples + other 7641 samples - 6550 negative samples)

手机上的澳门永利真的假的7196 images, 18159 bboxes

count_images_for_bbox:

count_bbox 1 2 3 4 5 6 7 8 9 10 11 12 13
count_images 2745 1922 958 568 441 234 83 85 104 21 13 21 1

count_bbox_min_size:

bbox_min_size 0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 360
count 144 7737 5079 2531 1385 673 312 144 86 35 18 11 1 1 1 1

average bbox_min_size: 42.3916876326

count_bbox_height_width_ratio:

bbox_height_width_ratio 0 1 2 3 4 5 6 7
count 5 16655 1353 116 19 9 1 1

手机上的澳门永利真的假的average bbox_height_width_ratio: 1.14734565069

count_images_for_category_id:

category_id 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
count_images 6 292 6 474 1077 0 0 0 0 54 1 12 13 4 3 15 254 9 176 57
category_id 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
count_images 93 19 0 371 172 55 11 112 2 269 69 1 9 47 235 978 111 4 29 4
category_id 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
count_images 14 23 28 89 2 2 31 162 1 37 485 84 2 0 1 11 2 14 47 0
category_id 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80
count_images 47 2 0 0 104 2 10 1 7 2 6 3 15 2 12 1 20 5 0 80
category_id 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
count_images 21 4 1 112 8 66 4 1 0 3 0 1 223 23 72 97 6 383 20 880
category_id 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120
count_images 681 518 1 100 542 44 269 39 159 0 2 5 0 13 8 47 2 25 105 9
category_id 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
count_images 72 4 7 6 9 95 0 1934 1 1415 0 549 1 27 47 136 0 30 42 16
category_id 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160
count_images 9 2 0 62 1 1 3 4 1 4 1 3 0 1 0 11 0 1 0 0
category_id 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180
count_images 7 0 4 0 1 98 0 20 32 0 10 0 11 46 46 0 12 0 9 0
category_id 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200
count_images 0 0 72 3 66 0 12 2 0 4 3 0 0 7 26 6 1 16 5 35
category_id 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
count_images 1 1 3 0 0 0 0 109 0 252 49 124 1 0 1 73 0 0 4 0
category_id 221
count_images 60

avg 75.1990950226 max 1934 (1) min 0 (46)

test

test: 3071 samples

3071 images, 8190 bboxes

count_images_for_bbox:

count_bbox 1 2 3 4 5 6 7 8 9 10 11 12 13 18
count_images 1042 844 407 278 236 111 53 39 42 6 4 6 2 1

count_bbox_min_size:

bbox_min_size 0 20 40 60 80 100 120 140 160 180 200 220 240 260 300 320 340 400
count 16 3440 2459 1107 561 306 139 77 44 23 7 2 2 2 2 1 1 1

手机上的澳门永利真的假的average bbox_min_size: 42.8355321502

count_bbox_height_width_ratio:

bbox_height_width_ratio 0 1 2 3 4 5
count 8 7402 699 69 10 2

手机上的澳门永利真的假的average bbox_height_width_ratio: 1.15697810748

手机上的澳门永利真的假的count_images_for_category_id:

category_id 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
count_images 1 140 2 223 461 0 0 0 0 15 2 1 4 1 3 3 103 5 84 17
category_id 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
count_images 39 2 0 175 114 15 5 58 2 117 39 0 4 14 86 491 66 2 13 1
category_id 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
count_images 6 9 12 33 0 0 8 99 0 12 230 47 1 0 0 0 0 11 13 0
category_id 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80
count_images 18 0 0 0 44 0 2 0 5 1 2 0 0 0 4 1 0 4 1 36
category_id 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
count_images 8 2 0 55 3 30 0 1 0 0 1 0 130 12 17 55 2 202 1 432
category_id 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120
count_images 338 262 0 44 259 3 124 3 79 0 0 2 1 2 1 1 0 5 47 0
category_id 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
count_images 31 1 1 0 2 37 0 913 0 615 0 282 0 5 2 63 2 6 10 2
category_id 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160
count_images 2 0 0 13 0 0 1 7 1 3 0 0 0 0 1 5 0 0 0 0
category_id 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180
count_images 0 0 0 0 1 31 0 2 3 0 3 0 3 9 14 0 3 0 0 0
category_id 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200
count_images 1 0 20 0 33 0 1 1 0 0 0 0 0 0 1 1 0 4 3 7
category_id 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
count_images 0 0 0 0 0 0 0 59 1 115 19 60 0 0 0 13 0 0 0 0
category_id 221
count_images 34

avg 33.8280542986 max 913 (1) min 0 (85)