Face mask detection using Inception Network Mask Classifier
Covvid-19 is here to stay for some time. While a lot of esteemed scientists are developing vaccine to fight the virus, we need to ensure that proper guuidelines are being adapted at various places to prevent the spread of virus. Keeping this intenton in mind, this face mask detector algorithm has been developed.
dataset credits : @prajnasb @Prajna Bhandary
Method adapted:
- Segregate directories into train and vaidation with sub-directories as "yes" or "no"
- Visualize that data has been split correctly
- Train model using pretrained Inception_v3 Model from Tensorflow Hub or Baseline model consisting of CNN and Max Pool Layers.
- Used the trained model to predict face mask detection usiing your Webcam.
Experimental Procedure:
- Utils scripts data_segment.py is imported in main_mask
- Baseline Model script baseline_mask.py is imported in main_mask
- Inception Model script inception_cnn is imported in main_mask
- Run main_mask.ipynb
Accuracy: Train accuracy: 99.6% Test accuracy: 98.4%
Applications:
Face mask detector can be used at various public places to ensure that people are practicing proper guidelines. Some of the public places include:
a. Public Transportation like Railways, Metros, buses etc. b. Schools/educational institutions c. Corporate Offices
Scope for Improvement:
a. Computation power can be reduced so as to make it more efficient with embedded devices with low computation power b. Dataset is artificially created and has masks only of a particular type, more real/artifical data can be added with different masks style and types(N-95 masks or colored masks etc.) c. In this dataset, samples of images with mask and without mask is of sample, this creates a identity bias which will definitely cause a problem in real world application, more data of people with no mask should be added to remove that bias.