Person Detection Model Zoo

Documentation Models
1

Dataset

Column 1 Column 2 Column 3 Column 4 E
Open Images Dataset 20,000 [1] 7,144 0: Person Annotations under CC BY 4.0, Images under CC BY 2.0

Model Zoo

DeGirum’s person detection model zoo features models trained with the Ultralytics repository.

Model Architecture Input Size Precision Runtime Device Type mAP 50-95 mAP 50 FPS
yolov5n_relu6_person 640x640 INT8 N2X ORCA1 0.2596 0.4709 153.1
yolov5n_relu6_person 640x640 INT8 OPENVINO CPU 0.2657 0.4742 –
yolov5n_relu6_person 640x640 FP32 OPENVINO CPU 0.2679 0.4782 –
yolov5nu_relu6_person 640x640 INT8 N2X ORCA1 0.2648 0.449 117.8
yolov5nu_relu6_person 640x640 INT8 TFLITE EDGETPU 0.2648 0.4495 –
yolov5nu_relu6_person 640x640 INT8 OPENVINO CPU 0.2629 0.4479 –
yolov5nu_relu6_person 640x640 FP32 OPENVINO CPU 0.2673 0.4534 –
yolov5nu_silu_person 640x640 INT8 N2X ORCA1 0.2698 0.454 –
yolov5nu_silu_person 640x640 INT8 TFLITE EDGETPU 0.2709 0.454 –
yolov5nu_silu_person 640x640 INT8 OPENVINO CPU 0.269 0.45 –
yolov5nu_silu_person 640x640 FP32 OPENVINO CPU 0.2753 0.4605 –
yolov8n_relu6_person 640x640 INT8 N2X ORCA1 0.2805 0.4682 107.3
yolov8n_relu6_person 640x640 INT8 TFLITE EDGETPU 0.2802 0.4685 –
yolov8n_relu6_person 640x640 INT8 OPENVINO CPU 0.2791 0.4665 –
yolov8n_relu6_person 640x640 FP32 OPENVINO CPU 0.282 0.4702 –
yolov8n_silu_person 640x640 INT8 N2X ORCA1 0.2671 0.4491 –
yolov8n_silu_person 640x640 INT8 TFLITE EDGETPU 0.2671 0.4471 –
yolov8n_silu_person 640x640 INT8 OPENVINO CPU 0.2719 0.4571 –
yolov8n_silu_person 640x640 FP32 OPENVINO CPU 0.2741 0.4597 –
yolov5s_relu6_person 640x640 INT8 N2X ORCA1 0.309 0.5317 77.1
yolov5s_relu6_person 640x640 INT8 OPENVINO CPU 0.3164 0.5355 –
yolov5s_relu6_person 640x640 FP32 OPENVINO CPU 0.3177 0.5367 –
yolov5su_relu6_person 640x640 INT8 N2X ORCA1 0.2988 0.4955 43.9
yolov5su_relu6_person 640x640 INT8 TFLITE EDGETPU 0.2986 0.495 –
yolov5su_relu6_person 640x640 INT8 OPENVINO CPU 0.2967 0.4956 –
yolov5su_relu6_person 640x640 FP32 OPENVINO CPU 0.2991 0.4969 –
yolov5su_silu_person 640x640 INT8 N2X ORCA1 0.2904 0.4847 –
yolov5su_silu_person 640x640 INT8 TFLITE EDGETPU 0.2908 0.4852 –
yolov5su_silu_person 640x640 INT8 OPENVINO CPU 0.2912 0.4843 –
yolov5su_silu_person 640x640 FP32 OPENVINO CPU 0.2947 0.4901 –
yolov8s_relu6_person 640x640 INT8 N2X ORCA1 0.3041 0.5022 38.9
yolov8s_relu6_person 640x640 INT8 TFLITE EDGETPU 0.3039 0.5016 –
yolov8s_relu6_person 640x640 INT8 OPENVINO CPU 0.3008 0.4986 –
yolov8s_relu6_person 640x640 FP32 OPENVINO CPU 0.3025 0.4998 –
yolov8s_silu_person 640x640 INT8 N2X ORCA1 0.2981 0.4932 –
yolov8s_silu_person 640x640 INT8 TFLITE EDGETPU 0.2966 0.4921 –
yolov8s_silu_person 640x640 INT8 OPENVINO CPU 0.3061 0.5047 –
yolov8s_silu_person 640x640 FP32 OPENVINO CPU 0.3052 0.5041 –
yolov5m_relu6_person 640x640 INT8 N2X ORCA1 0.3354 0.5578 24.4
yolov5m_relu6_person 640x640 INT8 OPENVINO CPU 0.3443 0.5595 –
yolov5m_relu6_person 640x640 FP32 OPENVINO CPU 0.3451 0.5623 –

  1. A subset of face training images are used for training. ↩︎