Deep learning quality practice case: Keras to achieve mask detection in the crowd

Posted May 28, 20202 min read

Keras is an open source artificial neural network library written in Python. It can be used as a high-level application program interface for Tensorflow, Microsoft-CNTK, and Theano to design, debug, evaluate, apply, and visualize deep learning models.

The excellent developer on Gitee is also a deep learning enthusiast. I learned the theory of RCNN, FastRCNN, and FasterRCNN in order to do a target detection. Then this project was born. Let's take a look at how he did it.

Project name: keras mask detection

Project Author: TANG ZHEN Ultra

Open source license agreement: GPL-3.0

Project address: https://gitee.com/tang \ _zhen \ _chao/keraskouzhaojiance

Introduction

Using keras to build fasterRCNN, training on the VOC format mask data set, to achieve the purpose of detecting whether the mask is worn in the crowd.

Software Architecture

\ * You can go to the project homepage to view

Installation tutorial

  1. Need numpy, matplotlib, scikit-learn, Pillow, tensorflow1.x, keras
  2. pip install package or conda install package
  3. If you want to train, you need to use VOC format data set, you need to install labelimg

Instructions for use

  1. There must be a model weight file under ./model_data/logs. Because the weight is large, it is not uploaded
  2. Under ./theory is the process of doing this thing and the explanation of the principle to facilitate understanding of the code
  3. Under ./net is the model construction of fasterRCNN
  4. run.py is used to directly run to see the results, voctrain.py is used to train your own VOC data set

The results show that

Click the link to go to the project homepage to see the project details: https://gitee.com/tang_zhen_chao/keraskouzhaojiance