|
|
|
|
This page covers the Tsinghua-Daimler Cyclist Detection Benchmark Dataset introduced in
|
|
|
|
X. Li, F. Flohr, Y. Yang, H. Xiong, M. Braun, S. Pan, K. Li and D. M. Gavrila. A New Benchmark for Vision-Based Cyclist Detection. In Proc. of the IEEE Intelligent Vehicles Symposium (IV), Gothenburg, Sweden, pp.1028-1033, 2016.
|
|
|
|
|
|
|
|
|
|
The Tsinghua-Daimler Cyclist Benchmark provides a benchmark dataset for cyclist detection. Bounding Box based labels are provided for the classes: ("pedestrian", "cyclist", "motorcyclist", "tricyclist", "wheelchairuser", "mopedrider").
The dataset consist of 4 subsets:
- Train Usually used for training, contains 9741 images with annotations only for "cyclist". Only cyclists which are fully visible (occlusion<10%) and higher than 60 pixels have been labeled here.
- Valid 1019 images to be used for validation of hyper parameters. Annotations for ("pedestrian", "cyclist", "motorcyclist", "tricyclist", "wheelchairuser", "mopedrider"). Only objects higher than 20 pixels have been labeled here.
- Test 2914 images normally used for testing with annotations for ("pedestrian", "cyclist", "motorcyclist", "tricyclist", "wheelchairuser", "mopedrider"). Only objects higher than 20 pixels have been labeled here.
- NonVRU 1000 images in which no object of interest ("pedestrian", "cyclist", "motorcyclist", "tricyclist", "wheelchairuser", "mopedrider") is present.
|
|
|
|
Dataset Statistics
|
Training
|
Validation
|
Test
|
Non-VRU
|
Total
|
# image frames
|
9741
|
1019
|
2914
|
1000
|
14674
|
# cyclist BBs
|
16202
|
1314
|
4657
|
0
|
22173
|
# pedestrian BBs
|
0 (ignored)
|
1541
|
7401
|
0
|
8942
|
# other rider BBs
|
0 (ignored)
|
190
|
1105
|
0
|
1295
|
# total BBs
|
16202
|
3045
|
13163
|
0
|
32410
|
|
|