blazeface

BlazeFace

Input

(Image from https://github.com/hollance/BlazeFace-PyTorch/blob/master/3faces.png)

Ailia input shape: (128, 128, 3)
Range: [-1, 1]

Output

Usage

Automatically downloads the tflite file on the first run. It is necessary to be connected to the Internet while downloading.

For the sample image,

$ python3 blazeface.py 

If you want to specify the input image, put the image path after the --input option.
You can use --savepath option to change the name of the output file to save.

$ python3 blazeface.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH

By adding the --video option, you can input the video.
If you pass 0 as an argument to VIDEO_PATH, you can use the webcam input instead of the video file.

$ python3 blazeface.py --video VIDEO_PATH --savepath SAVE_VIDEO_PATH

Two versions of the model are provided: full integer quantization (8-bit) and full precision floating point (32-bit). By default, the full integer quantization is used but the user can select the other version by passing the --float flag.

$ python3 blazeface.py --float

Reference

PINTO_model_zoo

Framework

TensorFlow 2.4.1, 1.15

Netron

• face_detection_front.tflite
• face_detection_front_128_full_integer_quant.tflite