[Retracted] A Review of Intelligent Driving Pedestrian Detection Based on Deep Learning

Table 3

Dataset related information.

Dataset name	Category	Images of train	Images of test	Size	Characteristic
Caltech	1	192000 (persons)	155000 (persons)	640 × 480	Large amount of data and rich annotation information
KITTI	8	7481	7518	1242 × 375	Capture datasets in rural areas and highways, each image contains up to 15 cars and 30 pedestrians
CityPersons	1	2975	500	2048 × 1024	The training set contains approximately 19,744 pedestrians and the test set contains 11 000 pedestrians
EuroCity	1	47300 (238300 persons)	—	1920 × 1024	Pedestrians and riders are carefully marked; especially posters and portraits are marked separately
TUD	1	1284	250	720 × 576	Evaluate the role of motion information in pedestrian detection and provide image pairs to calculate optical flow information
COCO	80	118000	46000	—	Multiple categories, large-scale datasets