Deep Learning to See

The remarkable progress in computer vision over the last few years is, by and large, attributed to deep learning, fueled availability of huge sets labeled data, paired with explosive growth GPU paradigm. While subscribing this view, book criticizes supposed scientific field proposes investigation within framework information-based laws nature. Specifically, present work poses fundamental questions about that remain far from understood, leading reader on a journey populated novel challenges resonating foundations machine learning. central thesis is for deeper understanding visual computational processes, it necessary look beyond applications general purpose learning algorithms focus instead appropriate theories take into account spatiotemporal nature signal.