Small object detection algorithm in UAV aerial images based on improved YOLO11

Table 1. Parameters and FLOPs of YOLO11 series model
View table
View in Article
Table 1. Parameters and FLOPs of YOLO11 series model
Model Input size/pixels Params/M FLOPs/B
YOLO11n 640 2.6 6.5
YOLO11s 640 9.4 21.5
YOLO11m 640 20.1 68.0
YOLO11l 640 25.3 86.9
YOLO11x 640 56.9 194.9

Table 2. Ablation experiment results of our algorithm modules on the VisDrone2021-val set
View table
View in Article
Table 2. Ablation experiment results of our algorithm modules on the VisDrone2021-val set
Baseline Tiny Head ACFI SPD-Conv AP_S/% AP_XS/% mAP_50-95/% Params/M FLOPs/B
√ 16.9 8.4 27.7 9.4 21.6
√ √ 20.6 11.5 30.8 9.6 29.6
√ √ 17.3 8.8 28.3 18.5 33.7
√ √ √ 21.0 11.8 31.5 19.0 46.4
√ √ √ √ 21.1 11.9 31.7 18.0 42.5

Table 3. Comparison results of different models on VisDrone2021

View table

View in Article

Table 3. Comparison results of different models on VisDrone2021

Model	mAP₅₀/%	AP_S/%	AP_XS/%	mAP_50-95/%	Params/M	FLOPs/B
Faster R-CNN^［7］	25.9	5.5	0.5	15.6	41.4	86.1
Retinanet^［10］	16.1	1.4	0.4	9.4	36.5	78.1
YOLOv5s	43.6	16.9	8.0	26.0	9.1	24.1
YOLOv8s	44.7	16.6	8.5	27.4	9.8	23.6
RT-DETR^［31］	40.9	13.5	6.3	24.4	32.8	108.0
YOLOv10s	42.7	15.8	7.7	25.8	8.1	24.8
YOLO11s	45.1	16.9	8.4	27.7	9.4	21.6
UAV-YOLOv8^［32］	49.6	20.1	10.9	30.4	5.3	55.4
HIC-YOLOv5^［33］	47.5	17.9	10.2	28.4	9.3	31.2
ARF-YOLOv8^［34］	47.8	20.9	11.7	29.4	5.4	86.1
ACFI-YOLO11	50.3	21.1	11.9	31.7	18.0	42.5

Table 4. mAP_50-95 and mAP₅₀ of each category of objects on the validation set of VisDrone2021

View table

View in Article

Table 4. mAP_50-95 and mAP₅₀ of each category of objects on the validation set of VisDrone2021

Model	Metrics	Pedestrian	People	Bicycle	Car	Van	Truck	Tricycle	Awning-tricycle	Bus	Motor	All
YOLO11s	mAP_50-95/%	24.5	16.1	9.1	61.0	36.3	27.9	18.0	12.0	46.9	24.8	27.7
YOLO11s	mAP₅₀/%	51.1	39.7	19.7	83.7	50.4	41.5	31.6	18.6	61.8	52.6	45.1
ACFI-YOLO11	mAP_50-95/%	30.1	20.8	12.2	64.7	40.4	30.3	22.3	14.5	51.5	30.4	31.7
ACFI-YOLO11	mAP₅₀/%	59.3	47.5	25.4	87.1	54.4	44.3	36.6	22.1	66.6	60.0	50.3

Table 5. mAP_50-95 of each category of objects on the validation set of VisDrone2021
View table
View in Article
Table 5. mAP_50-95 of each category of objects on the validation set of VisDrone2021
Model Car Truck Bus All
RT-DETR 69.0 74.0 79.7 74.2
HIC-YOLOv5 73.3 76.4 80.1 76.6
UAV-YOLOv8 78.6 82.4 86.2 82.4
YOLO11s 77.0 80.6 85.4 81.0
ACFI-YOLO11 79.5 83.3 87.0 83.3

Tools

Get Citation

Copy Citation Text

Zhihao ZHANG, Xiaorun LI, Shuhan CHEN. Small object detection algorithm in UAV aerial images based on improved YOLO11[J]. Chinese Journal of Liquid Crystals and Displays, 2025, 40(6): 915

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category:

Received: Jan. 10, 2025

Accepted: --

Published Online: Jul. 14, 2025

The Author Email: Xiaorun LI (lxr@zju.edu.cn)

DOI:10.37188/CJLCD.2025-0010

Topics

Table 1. Parameters and FLOPs of YOLO11 series model

Table 1. Parameters and FLOPs of YOLO11 series model

Table 2. Ablation experiment results of our algorithm modules on the VisDrone2021-val set

Table 2. Ablation experiment results of our algorithm modules on the VisDrone2021-val set

Table 3. Comparison results of different models on VisDrone2021

Table 3. Comparison results of different models on VisDrone2021

Table 4. mAP50-95 and mAP50 of each category of objects on the validation set of VisDrone2021

Table 4. mAP50-95 and mAP50 of each category of objects on the validation set of VisDrone2021

Table 5. mAP50-95 of each category of objects on the validation set of VisDrone2021

Table 5. mAP50-95 of each category of objects on the validation set of VisDrone2021

Table 4. mAP_50-95 and mAP₅₀ of each category of objects on the validation set of VisDrone2021

Table 4. mAP_50-95 and mAP₅₀ of each category of objects on the validation set of VisDrone2021

Table 5. mAP_50-95 of each category of objects on the validation set of VisDrone2021

Table 5. mAP_50-95 of each category of objects on the validation set of VisDrone2021