VisDrone Detection Model Zoo

Documentation Models
1

Dataset

Dataset Name Training Images Validation Images Class Labels License
VisDrone 2019 Dataset 6,471 548 10 classes CC BY-NC-SA 3.0

An example of each class is given from left-to-right:

Class ID Label Description
0 Pedestrian A person who is standing/walking.
1 People A person who is not standing/walking.
2 Bicycle A bicycle.
3 Car A car around the size of a sedan.
4 Van A van.
5 Truck A vehicle with an open load bed.
6 Tricycle A three wheeled vehicle, pedal or motor operated.
7 Awning Tricycle A three wheeled vehicle, with a roof of some sorts.
8 Bus A bus.
9 Motor A moped or motorcycle.

visdrone_classes
visdrone_classes858×353 249 KB

Model Zoo

DeGirum’s VisDrone detection model zoo offers models trained with the Ultralytics repository.

Model Architecture Input Size Precision Runtime Device Type mAP 50-95 mAP 50 mAP 50-95 Small mAP 50-95 Medium mAP 50-95 Large FPS
yolov8n_relu6_visdrone 640x384 INT8 N2X ORCA 0.143 0.2521 0.3226 0.2287 0.3226 154.6
yolov8n_relu6_visdrone 640x384 INT8 TFLITE EDGETPU 0.1427 0.2516 0.3262 0.2279 0.3261 –
yolov8n_relu6_visdrone 640x384 INT8 OPENVINO CPU 0.1442 0.2531 0.3255 0.2284 0.3255 –
yolov8n_relu6_visdrone 640x384 FP32 OPENVINO CPU 0.1454 0.2552 0.3312 0.2295 0.3312 –
yolov8n_relu6_visdrone 960x544 INT8 N2X ORCA 0.1963 0.3349 0.3575 0.2925 0.3575 86.6
yolov8n_relu6_visdrone 960x544 INT8 TFLITE EDGETPU 0.1968 0.3363 0.3554 0.2935 0.3553 –
yolov8n_relu6_visdrone 960x544 INT8 OPENVINO CPU 0.2015 0.3388 0.3606 0.3018 0.3606 –
yolov8n_relu6_visdrone 960x544 FP32 OPENVINO CPU 0.2027 0.3405 0.3606 0.3036 0.3606 –
yolov8n_relu6_visdrone 1280x736 INT8 N2X ORCA 0.2328 0.3948 0.3787 0.3294 0.3786 58.6
yolov8n_relu6_visdrone 1280x736 INT8 TFLITE EDGETPU 0.2331 0.395 0.3771 0.3306 0.3770 –
yolov8n_relu6_visdrone 1280x736 INT8 OPENVINO CPU 0.2393 0.3981 0.3838 0.3424 0.3838 –
yolov8n_relu6_visdrone 1280x736 FP32 OPENVINO CPU 0.2406 0.3999 0.3884 0.3425 0.3883 –
yolov8s_relu6_visdrone 640x384 INT8 N2X ORCA 0.1874 0.3193 0.3938 0.2912 0.3938 56.8
yolov8s_relu6_visdrone 640x384 INT8 TFLITE EDGETPU 0.1874 0.3187 0.3884 0.2918 0.3883 –
yolov8s_relu6_visdrone 640x384 INT8 OPENVINO CPU 0.1903 0.3219 0.3847 0.2955 0.3846 –
yolov8s_relu6_visdrone 640x384 FP32 OPENVINO CPU 0.1904 0.3223 0.3889 0.2950 0.3888 –
yolov8s_relu6_visdrone 960x544 INT8 N2X ORCA 0.249 0.417 0.4409 0.3647 0.4408 28.3
yolov8s_relu6_visdrone 960x544 INT8 TFLITE EDGETPU 0.2496 0.4172 0.4402 0.3658 0.4401 –
yolov8s_relu6_visdrone 960x544 INT8 OPENVINO CPU 0.2508 0.4188 0.4371 0.3683 0.4371 –
yolov8s_relu6_visdrone 960x544 FP32 OPENVINO CPU 0.2515 0.4188 0.4394 0.3692 0.4393 –
yolov8s_relu6_visdrone 1280x736 INT8 N2X ORCA 0.2863 0.4735 0.4572 0.3946 0.4572 17.8
yolov8s_relu6_visdrone 1280x736 INT8 TFLITE EDGETPU 0.2867 0.4737 0.457 0.3942 0.4569 –
yolov8s_relu6_visdrone 1280x736 INT8 OPENVINO CPU 0.293 0.4765 0.4639 0.4071 0.4638 –
yolov8s_relu6_visdrone 1280x736 FP32 OPENVINO CPU 0.2942 0.4776 0.4696 0.4076 0.4696 –
yolov8m_relu6_visdrone 640x384 INT8 N2X ORCA 0.2111 0.3557 0.4453 0.3249 0.4453 24.5
yolov8m_relu6_visdrone 640x384 INT8 OPENVINO CPU 0.2156 0.3595 0.4415 0.3316 0.4415 –
yolov8m_relu6_visdrone 640x384 FP32 OPENVINO CPU 0.2155 0.3589 0.4483 0.3310 0.4483 –
yolov8m_relu6_visdrone 960x544 INT8 N2X ORCA 0.2812 0.4602 0.4731 0.4006 0.4730 2.1
yolov8m_relu6_visdrone 960x544 INT8 OPENVINO CPU 0.2864 0.4634 0.4813 0.4091 0.4812 –
yolov8m_relu6_visdrone 960x544 FP32 OPENVINO CPU 0.2876 0.4639 0.4877 0.4095 0.4877 –
yolov8m_relu6_visdrone 1280x736 INT8 N2X ORCA 0.3048 0.5078 0.4639 0.4141 0.4639 0.7
yolov8m_relu6_visdrone 1280x736 INT8 OPENVINO CPU 0.3212 0.518 0.4662 0.4383 0.4662 –
yolov8m_relu6_visdrone 1280x736 FP32 OPENVINO CPU 0.3219 0.5189 0.4681 0.4393 0.4680 –
yolov8l_relu6_visdrone 640x384 INT8 N2X ORCA 0.226 0.3775 0.4664 0.3475 0.4663 13.0
yolov8l_relu6_visdrone 640x384 INT8 OPENVINO CPU 0.232 0.3824 0.4812 0.3556 0.4811 –