Object Detection With Deep Learning: A Review

This paper is a review of object detection frameworks based on deep learning. It contrasts modern approaches with traditional detection methods that rely on handcrafted features and shallow trainable architectures, then introduces the history of deep learning and its representative tool, the convolutional neural network. The review focuses on typical generic object detection architectures and describes modifications and useful tricks that further improve detection performance.

Because different detection tasks have different characteristics, the authors also survey several specific ones, including salient object detection, face detection, and pedestrian detection, and provide experimental analyses that compare various methods and draw conclusions. By consolidating architectures, tricks, task-specific insights, and experimental comparisons in a single reference, the review offers guidelines intended to steer future work in object detection and related neural network-based learning systems.