Gender Classification
Dataset
| Dataset Name | Training Images | Validation Images | Class Labels | License |
|---|---|---|---|---|
| UTKFace CelebA FairFace |
249,575 | 27,652 | 0: Male 1: Female |
UTKFace: Non-commercial research purposes only CelebA: Non-commercial research purposes only FairFace: CC BY 4.0 |
Model Zoo
DeGirum's gender classification model (Female | Male) zoo consists of a variety of models gathered and trained using ultralytics repository and pytorch-image-models repository.
Here is the link to DeGirum's gender classification zoo.
| Model Architecture | Input Size | Precision | Runtime | Device Type | Top-12 | FPS |
|---|---|---|---|---|---|---|
| yolov8n_relu6_gender | 160x160 | INT8 | N2X | ORCA1 | 0.9558 | 35.9 |
| yolov8n_relu6_gender | 160x160 | INT8 | RKNN | RK3588 | 0.9626 | -- |
| yolov8n_relu6_gender | 160x160 | INT8 | OPENVINO | CPU | 0.9578 | -- |
| yolov8n_relu6_gender | 160x160 | FP32 | OPENVINO | CPU | 0.9629 | -- |
| yolov8n_relu6_gender | 256x256 | INT8 | N2X | ORCA1 | 0.9686 | 31.3 |
| yolov8n_relu6_gender | 256x256 | INT8 | RKNN | RK3588 | 0.9646 | -- |
| yolov8n_relu6_gender | 256x256 | INT8 | OPENVINO | CPU | 0.9625 | -- |
| yolov8n_relu6_gender | 256x256 | FP32 | OPENVINO | CPU | 0.9654 | -- |
| yolov8s_relu6_gender | 160x160 | INT8 | N2X | ORCA1 | 0.9670 | 31.4 |
| yolov8s_relu6_gender | 160x160 | INT8 | RKNN | RK3588 | 0.9686 | -- |
| yolov8s_relu6_gender | 160x160 | INT8 | OPENVINO | CPU | 0.9663 | -- |
| yolov8s_relu6_gender | 160x160 | FP32 | OPENVINO | CPU | 0.9670 | -- |
| yolov8s_relu6_gender | 256x256 | INT8 | N2X | ORCA1 | 0.9702 | 25.3 |
| yolov8s_relu6_gender | 256x256 | INT8 | RKNN | RK3588 | 0.9693 | -- |
| yolov8s_relu6_gender | 256x256 | INT8 | OPENVINO | CPU | 0.9693 | -- |
| yolov8s_relu6_gender | 256x256 | FP32 | OPENVINO | CPU | 0.9697 | -- |
| mobilenetv2_100_gender | 160x160 | INT8 | N2X | ORCA1 | 0.8616 | 285 |
| mobilenetv2_100_gender | 224x224 | INT8 | N2X | ORCA1 | 0.958 | 220 |
| mobilenetv2_100_gender | 224x224 | FP32 | OPENVINO | CPU | 0.9648 | -- |
| mobilenetv2_100_gender | 256x256 | INT8 | N2X | ORCA1 | 0.9579 | 203.6 |
| resnet_34_gender | 160x160 | INT8 | N2X | ORCA1 | 0.843 | 91.6 |
| resnet_34_gender | 160x160 | FP32 | OPENVINO | CPU | 0.8433 | -- |
| resnet_34_gender | 224x224 | INT8 | N2X | ORCA1 | 0.9633 | 77.9 |
| resnet_34_gender | 224x224 | INT8 | OPENVINO | CPU | 0.9628 | -- |
| resnet_34_gender | 256x256 | INT8 | N2X | ORCA1 | 0.9635 | 75 |
| resnet_34_gender | 256x256 | INT8 | OPENVINO | CPU | 0.9642 | -- |
| resnet_34_gender | 256x256 | FP32 | OPENVINO | CPU | 0.9645 | -- |
| resnet_50_gender | 160x160 | INT8 | N2X | ORCA1 | 0.9182 | 71.9 |
| resnet_50_gender | 160x160 | FP32 | OPENVINO | CPU | 0.9013 | -- |
| resnet_50_gender | 224x224 | INT8 | N2X | ORCA1 | 0.9603 | 61.8 |
| resnet_50_gender | 256x256 | FP32 | OPENVINO | CPU | 0.9616 | 58.8 |
2: Evaluation of models was performed on original images without loss due to JPEG compression.