[AI] "Hands-on Learning-Deep Learning-PyTorch Edition" Notes (22): Single-shot Multi-frame Detection (SSD)

Summary of AI learning catalog

1 Introduction

SSD (Single Shot MultiBox Detector) single shot multi-frame detection. "Single shot" indicates that the SSD algorithm is a one-stage (one-stage) method, and "MultiBox" indicates that SSD is a multi-frame prediction (multi-scale anchor box/feature map).

SSD and YOLO both use CNN network to perform one-stage detection. The difference is:

• YOLO is fast and suitable for real-time object detection; but its accuracy is slightly lower, especially for the detection of small objects.
• SSD has high accuracy, especially when detecting objects of various sizes at the same time; but the speed is slightly lower.

The following lists one-stage (one-stage) and two-stage (two-stage) algorithms and their corresponding speed and accuracy.

2. What is one-stage?

One-stage is relative to two-stage. Let’s first understand the two-stage.