This model is designed for face detection in real time. This lightweight model is developed specifically for edge computing devices, and smartphones. Input tensor is a color image of size 1 x 3 x height x width with mean values of 127, 127, 127. Input images should be in the RGB format and resized to 320 x 240 pixels for this version of the model. The model outputs two arrays for scores (1 x 4420 x 2) and boxes (1 x 4420 x 4).
ZTN model