This tutorial aims to review computer vision techniques before and after the deep learning era, in critical domains such as object detection, texture classification, scene understanding and instance retrieval. In the computer vision community, dramatic evolution is witnessed in the past 25 years, especially in visual recognition. The tremendous success cannot be made possible without the development of feature representation and learning approaches, which are at the core of many visual recognition problems such as texture recognition, image classification, object detection and recognition, scene classification and content based instance retrieval.

In specific, we will focus on four closely related visual recognition problems at different levels: objects detection and recognition, scene understanding, and content based image retrieval. These problems have received significant attention from both academia and industry in the field of computer vision and pattern recognition. For each problem, this tutorial will firstly review the milestones in the two development stages, then present an overview of the current frontier and state of the art performance on leading benchmark datasets, and finally discuss the possible future research directions.