Multi-scale pedestrian detection algorithm based on joint head and overall information

算法名称	AP	MR^-2
Faster RCNN+FPN （Baseline）	84.2	50.2
PS-RCNN^［25］	86.0	—
RetinaNet^［26］	80.8	63.3
YOLOv3^［27］	83.1	54.2
YOLOv5^［28］	84.1	51.8
YOLO-CS^［29］	81.9	59.0
FCOS^［30］	—	48.3
Adaptive NMS^［31］	84.7	49.7
Soft-NMS^［31］	83.9	52.0
JED^［32］	85.9	—
YOLOv8	85.6	—
Ours	87.1	46.2

Table 2. Results of comparative experiments on the TJU-DHD-pedestrian dataset

View table

View in Article

Table 2. Results of comparative experiments on the TJU-DHD-pedestrian dataset

算法名称	TJU-Ped-campus子集		TJU-Ped-traffic子集
算法名称	Reasonable MR^-2	Reasonable Small MR^-2	ReasonableMR^-2	Reasonable SmallMR^-2
Faster R-CNN+FPN （Baseline）	28.5	73.9	22.7	35.8
RetinaNet^［26］	34.7	83.0	23.9	37.9
FCOS^［30］	31.9	81.3	24.4	37.4
Faster R-CNN+PANet^［33］	28.2	73.4	22.4	35.5
Ours	27.0	71.5	21.3	33.6

Table 3. Results of comparative experiments on the CityPersons dataset

View table

View in Article

Table 3. Results of comparative experiments on the CityPersons dataset

算法名称	Bare MR^-2	Reasonable MR^-2	Partial MR^-2	Heavy MR^-2
OR-CNN^［34］	6.7	12.8	15.3	55.7
MFGP^［35］	7.9	10.9	10.9	49.9
TLL^［36］	10.0	15.5	17.2	53.6
CCFA-Net^［6］	—	15.4	—	52.9
ALFNet^［37］	8.4	12.0	11.4	51.9
Repulsion Loss^［38］	7.6	13.2	16.8	56.9
Ours	7.4	10.6	10.5	47.8

Table 4. Results of ablation experiment
View table
View in Article
Table 4. Results of ablation experiment
算法名称循环递归特征金字塔网络 CA注意力机制行人头部与整体联合检测模块 AP MR^-2
本文算法 √ √ √ 87.1 46.2
√ √ × 85.5 48.5
√ × × 85.1 49.1
Baseline × × × 84.2 50.2

Tools

Get Citation

Copy Citation Text

Ximing MA, Ning LI, Di WU, Jianfei WANG, Xiangyue YU. Multi-scale pedestrian detection algorithm based on joint head and overall information[J]. Chinese Journal of Liquid Crystals and Displays, 2025, 40(6): 931

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites