Our project aims to create an integrated system combined with mask recognition and infrared thermometer. The system is able to tell the difference whether an individual is wearing a mask or not. Moreover, it can measure an individual's temperature at the same time. The camera and infrared probe will be utilized to implement our project.
- Tensorflow 2.3.0
- Python 3.8
- OpenCV 4.5
- Raspberry Pi 4B+
In this part, MLX90614 was used as the infrared body surface temperature detection module.
In this part, I used Tensorflow2.3.0 to construct my model, specificly, I used MobileNet and Transfer Learning.
2021.4.11
Discussed the basic timetable and the hardware platform we need(Raspberry Pi). In addition, I decided to use OpenCV to recongnize the human face.
2021.4.28
Installed TensorFlow and MiniConda. Tried simple models implemented in TensorFlow. Machine Learning studying is still in progress.
2021.5.13 - 2021.5.27
Choose the model and technique. I decided to use MobileNet and Transfer Learning to construct my model.
Then use the dataset(https://github.com/manish-1305/facemask_detection) to train my model.
Save it as .h5 file, make sure that I can use it later on.
Noted that the directories of dataset should organized as follows:
-Project directory
|-- Dataset
|---|-- No_Mask
|---|-- With_Mask
|-- train_model.py
Noted that the directory and file should organized as follows:
-Project directory
|-- Dataset
|---|-- No_Mask
|---|-- With_Mask
|-- train_model.py
|-- test_model.py
|-- haarcascade_frontalface_default.xml
[1]Howard A G, Zhu M, Chen B, et al. Mobilenets: Efficient convolutional neural networks for mobile vision applications[J]. arXiv preprint arXiv:1704.04861, 2017.
[2]Pan S J, Yang Q. A survey on transfer learning[J]. IEEE Transactions on knowledge and data engineering, 2009, 22(10): 1345-1359.
[3]https://hackcv.com/posts/%E7%BF%BB%E8%AF%91-%E7%A5%9E%E7%BB%8F%E7%BD%91%E7%BB%9C%E7%9A%84%E7%9B%B4%E8%A7%82%E8%A7%A3%E9%87%8A/