Slim-CNN: A Light-Weight CNN for Face Attribute Prediction
This work addresses the problem of efficient face attribute prediction for mobile and embedded applications, representing an incremental improvement in model compression.
The paper tackles face attribute prediction by introducing Slim-CNN, a lightweight neural network that achieves 91.24% accuracy on the CelebA dataset with at least 25 times fewer parameters and an 87% reduction in memory storage compared to similar methods.
We introduce a computationally-efficient CNN micro-architecture Slim Module to design a lightweight deep neural network Slim-Net for face attribute prediction. Slim Modules are constructed by assembling depthwise separable convolutions with pointwise convolution to produce a computationally efficient module. The problem of facial attribute prediction is challenging because of the large variations in pose, background, illumination, and dataset imbalance. We stack these Slim Modules to devise a compact CNN which still maintains very high accuracy. Additionally, the neural network has a very low memory footprint which makes it suitable for mobile and embedded applications. Experiments on the CelebA dataset show that Slim-Net achieves an accuracy of 91.24% with at least 25 times fewer parameters than comparably performing methods, which reduces the memory storage requirement of Slim-net by at least 87%.