Update README.md
Browse files
README.md
CHANGED
@@ -32,8 +32,9 @@ dataset_info:
|
|
32 |
'22': red and white triangle rough / bumpy road warning
|
33 |
'23': red and white triangle car skidding / slipping warning
|
34 |
'24': red and white triangle with merging / narrow lanes warning
|
35 |
-
'25':
|
36 |
-
|
|
|
37 |
'26': red and white triangle with traffic light approaching warning
|
38 |
'27': red and white triangle with person walking warning
|
39 |
'28': red and white triangle with child and person walking warning
|
@@ -49,9 +50,11 @@ dataset_info:
|
|
49 |
'38': blue circle with white keep right arrow mandatory
|
50 |
'39': blue circle with white keep left arrow mandatory
|
51 |
'40': blue circle with white arrows indicating a traffic circle
|
52 |
-
'41':
|
|
|
53 |
ended
|
54 |
-
'42':
|
|
|
55 |
has ended
|
56 |
splits:
|
57 |
- name: train
|
@@ -83,4 +86,135 @@ dataset_info:
|
|
83 |
num_examples: 12630
|
84 |
download_size: 841108239
|
85 |
dataset_size: 15718682
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
32 |
'22': red and white triangle rough / bumpy road warning
|
33 |
'23': red and white triangle car skidding / slipping warning
|
34 |
'24': red and white triangle with merging / narrow lanes warning
|
35 |
+
'25': >-
|
36 |
+
red and white triangle with person digging / construction / road
|
37 |
+
work warning
|
38 |
'26': red and white triangle with traffic light approaching warning
|
39 |
'27': red and white triangle with person walking warning
|
40 |
'28': red and white triangle with child and person walking warning
|
|
|
50 |
'38': blue circle with white keep right arrow mandatory
|
51 |
'39': blue circle with white keep left arrow mandatory
|
52 |
'40': blue circle with white arrows indicating a traffic circle
|
53 |
+
'41': >-
|
54 |
+
white circle with gray strike bar indicating no passing for cars has
|
55 |
ended
|
56 |
+
'42': >-
|
57 |
+
white circle with gray strike bar indicating no passing for trucks
|
58 |
has ended
|
59 |
splits:
|
60 |
- name: train
|
|
|
86 |
num_examples: 12630
|
87 |
download_size: 841108239
|
88 |
dataset_size: 15718682
|
89 |
+
task_categories:
|
90 |
+
- image-classification
|
91 |
+
language:
|
92 |
+
- en
|
93 |
+
size_categories:
|
94 |
+
- 10K<n<100K
|
95 |
---
|
96 |
+
|
97 |
+
|
98 |
+
|
99 |
+
# Dataset Card for German Traffic Sign Recognition Benchmark
|
100 |
+
|
101 |
+
This dataset contains images of 43 classes of traffic signs. It is intended for developing and benchmarking traffic sign recognition systems.
|
102 |
+
|
103 |
+
## Dataset Details
|
104 |
+
|
105 |
+
### Dataset Description
|
106 |
+
|
107 |
+
The German Traffic Sign Recognition Benchmark (GTSRB) is a multi-class classification dataset featuring 43 classes of traffic signs.
|
108 |
+
The images were cropped from a larger set of images to focus on the traffic sign and eliminate background.
|
109 |
+
Multiple data augmentations such as Gaussian noise, motion blur, contrast changes, etc. are provided as additional test sets to benchmark model robustness.
|
110 |
+
|
111 |
+
### Dataset Sources
|
112 |
+
|
113 |
+
- [Paper with code](https://paperswithcode.com/dataset/gtsrb)
|
114 |
+
|
115 |
+
## Uses
|
116 |
+
|
117 |
+
### Direct Use
|
118 |
+
|
119 |
+
```python
|
120 |
+
from datasets import load_dataset
|
121 |
+
|
122 |
+
dataset = load_dataset('tanganke/gtsrb')
|
123 |
+
```
|
124 |
+
|
125 |
+
## Dataset Structure
|
126 |
+
|
127 |
+
The dataset is provided in 10 splits, including training data and clean test data:
|
128 |
+
|
129 |
+
- train: 26,640 images
|
130 |
+
- test: 12,630 images
|
131 |
+
|
132 |
+
and 7 kinds of corrupted test datasets to evaluate the robustness:
|
133 |
+
|
134 |
+
- contrast: 12,630 contrast-adjusted test images
|
135 |
+
- gaussian_noise: 12,630 Gaussian noise augmented test images
|
136 |
+
- impulse_noise: 12,630 impulse noise augmented test images
|
137 |
+
- jpeg_compression: 12,630 JPEG-compressed test images
|
138 |
+
- motion_blur: 12,630 motion-blurred test images
|
139 |
+
- pixelate: 12,630 pixelated test images
|
140 |
+
- spatter: 12,630 spatter augmented test images
|
141 |
+
|
142 |
+
Each split contains 43 classes of traffic signs, with the class labels and names specified in the dataset metadata.
|
143 |
+
|
144 |
+
## Citation [optional]
|
145 |
+
|
146 |
+
You can use any of the provided BibTeX entries for your reference list:
|
147 |
+
|
148 |
+
```bibtex
|
149 |
+
@article{stallkampManVsComputer2012,
|
150 |
+
title = {Man vs. Computer: {{Benchmarking}} Machine Learning Algorithms for Traffic Sign Recognition},
|
151 |
+
shorttitle = {Man vs. Computer},
|
152 |
+
author = {Stallkamp, J. and Schlipsing, M. and Salmen, J. and Igel, C.},
|
153 |
+
year = {2012},
|
154 |
+
month = aug,
|
155 |
+
journal = {Neural Networks},
|
156 |
+
series = {Selected {{Papers}} from {{IJCNN}} 2011},
|
157 |
+
volume = {32},
|
158 |
+
pages = {323--332},
|
159 |
+
issn = {0893-6080},
|
160 |
+
doi = {10.1016/j.neunet.2012.02.016},
|
161 |
+
url = {https://www.sciencedirect.com/science/article/pii/S0893608012000457},
|
162 |
+
keywords = {Benchmarking,Convolutional neural networks,Machine learning,Traffic sign recognition}
|
163 |
+
}
|
164 |
+
|
165 |
+
@misc{yangAdaMergingAdaptiveModel2023,
|
166 |
+
title = {{{AdaMerging}}: {{Adaptive Model Merging}} for {{Multi-Task Learning}}},
|
167 |
+
shorttitle = {{{AdaMerging}}},
|
168 |
+
author = {Yang, Enneng and Wang, Zhenyi and Shen, Li and Liu, Shiwei and Guo, Guibing and Wang, Xingwei and Tao, Dacheng},
|
169 |
+
year = {2023},
|
170 |
+
month = oct,
|
171 |
+
number = {arXiv:2310.02575},
|
172 |
+
eprint = {2310.02575},
|
173 |
+
primaryclass = {cs},
|
174 |
+
publisher = {arXiv},
|
175 |
+
doi = {10.48550/arXiv.2310.02575},
|
176 |
+
url = {http://arxiv.org/abs/2310.02575},
|
177 |
+
archiveprefix = {arxiv},
|
178 |
+
keywords = {Computer Science - Computer Vision and Pattern Recognition,Computer Science - Machine Learning}
|
179 |
+
}
|
180 |
+
|
181 |
+
@misc{tangConcreteSubspaceLearning2023,
|
182 |
+
title = {Concrete {{Subspace Learning}} Based {{Interference Elimination}} for {{Multi-task Model Fusion}}},
|
183 |
+
author = {Tang, Anke and Shen, Li and Luo, Yong and Ding, Liang and Hu, Han and Du, Bo and Tao, Dacheng},
|
184 |
+
year = {2023},
|
185 |
+
month = dec,
|
186 |
+
number = {arXiv:2312.06173},
|
187 |
+
eprint = {2312.06173},
|
188 |
+
publisher = {arXiv},
|
189 |
+
url = {http://arxiv.org/abs/2312.06173},
|
190 |
+
archiveprefix = {arxiv},
|
191 |
+
copyright = {All rights reserved},
|
192 |
+
keywords = {Computer Science - Machine Learning}
|
193 |
+
}
|
194 |
+
|
195 |
+
|
196 |
+
@misc{tangMergingMultiTaskModels2024,
|
197 |
+
title = {Merging {{Multi-Task Models}} via {{Weight-Ensembling Mixture}} of {{Experts}}},
|
198 |
+
author = {Tang, Anke and Shen, Li and Luo, Yong and Yin, Nan and Zhang, Lefei and Tao, Dacheng},
|
199 |
+
year = {2024},
|
200 |
+
month = feb,
|
201 |
+
number = {arXiv:2402.00433},
|
202 |
+
eprint = {2402.00433},
|
203 |
+
primaryclass = {cs},
|
204 |
+
publisher = {arXiv},
|
205 |
+
doi = {10.48550/arXiv.2402.00433},
|
206 |
+
url = {http://arxiv.org/abs/2402.00433},
|
207 |
+
archiveprefix = {arxiv},
|
208 |
+
copyright = {All rights reserved},
|
209 |
+
keywords = {Computer Science - Computer Vision and Pattern Recognition,Computer Science - Machine Learning}
|
210 |
+
}
|
211 |
+
```
|
212 |
+
|
213 |
+
|
214 |
+
## Dataset Card Authors
|
215 |
+
|
216 |
+
Anke Tang
|
217 |
+
|
218 |
+
## Dataset Card Contact
|
219 |
+
|
220 |
+
[tang.anke@foxmail.com](mailto:tang.anke@foxmail.com)
|