Cascaded Projection: End-to-End Network Compression and Acceleration

Breton Minnehan, Andreas Savakis

We propose a data-driven approach for deep convolutional neural network compression that achieves high accuracy with high throughput and low memory requirements. Current network co…