Shulou(Shulou.com)06/02 Report--

This article shows you how to use deepsort to achieve multi-target tracking in CenterNet, the content is concise and easy to understand, it can definitely brighten your eyes. I hope you can get something through the detailed introduction of this article.

Introduction to CenterNet

The traditional target detection methods based on key points, such as the most representative CornerNet, determine the target by detecting the upper left corner and the lower right corner of the object, but in the process of determining the target, the internal characteristics of the object can not be effectively used, that is, the internal information of the object can not be sensed, which leads to a lot of false detection (wrong target box). CenterNet uses the triple of key points, namely, the center point, the upper left corner and the lower right corner, instead of two key points to determine a target, so that the network has the ability to perceive the internal information of the object at a small cost, so that it can effectively restrain false detection. In addition, in order to better detect the center point and corner point, we propose center pooling and cascade corner pooling to extract the features of the center point and corner point respectively. The name of our method is CenterNet, which is an one-stage method.

The principle of error detection suppression is based on the following inference: if the target box is accurate, then the probability of detecting the target center point in its central area will be very high, and vice versa. Therefore, we first use the upper left and lower right corners to generate the initial target box, define a central area for each prediction box, and then determine whether the central area of each target box contains a central point, and if so, keep the target box. if not, delete the target box, the principle of which is shown in the following figure:

Code connection: https://github.com/xingyizhou/CenterNet

Introduction to Deepsort

Deepsort mainly consists of the following algorithms:

1. Kalman filter

2. Mahalanobis distance

3. PCA principal component analysis.

4. Hungarian algorithm

5. Pedestrian re-identification

6. MOT evaluation index

Each of them is a lot to talk about, so save time to explain it in detail later.

The following figure outlines and well demonstrates deepsort's algorithm:

Get the code

Git clone https://github.com/kimyoon-young/centerNet-deep-sort.git

Install repo

Conda env create-f CenterNet.yml

Pip install-r requirments.txt

Quick start

CENTERNET_PATH = 'CENTERNET_ROOT/CenterNet/src/lib/'

To

E.g) CENTERNET_PATH ='/ home/kyy/centerNet-deep-sort/CenterNet/src/lib/'

Run demo

The above content of python demo_centernet_deepsort.py is how to use deepsort to achieve multi-target tracking in CenterNet. Have you learned any knowledge or skills? If you want to learn more skills or enrich your knowledge reserve, you are welcome to follow the industry information channel.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.