aplux/Lightweight-Face-Detection

Lightweight-Face-Detection: Object Detection

Lightweight-Face-Detection is a mobile/edge-optimized face detection model utilizing depthwise separable convolutions and multi-scale feature fusion for high-accuracy real-time detection under low resources. Built on streamlined backbones (e.g., ShuffleNetV2) with dynamic anchor adjustment, it robustly detects small, occluded, or blurred faces, achieving 50+ FPS on mobile CPUs via multi-thread processing. Quantization-aware training (INT8) reduces model size below 2MB, reaching over 90% mAP on WIDER FACE—15% higher than MTCNN with faster speed. Ideal for smart access control, mobile camera focus, and video conferencing, it balances low power consumption and high performance.

Source model

• Input shape: 1x1x480x640
• Number of parameters: 0.84M
• Model size: 3.72M
• Output shape: [1x1x60x80],[1x4x60x80],[1x10x60x80]

The source model can be found here

Performance Reference

Please search model by model name in Model Farm

Inference & Model Conversion

Please search model by model name in Model Farm

License

• Source Model: BSD-3-CLAUSE

• Deployable Model: APLUX-MODEL-FARM-LICENSE