Discussion
Problems Encountered Potential Future Work
Currently, in our project, we don’t have enough dataset about people wearing masks improperly. We just detect people’s noses to classify the wearing mask whether properly or not. In the future, we can add more datasets and use neural networks to train our model directly, and use the model directly to detect wearing masks whether properly or not.
On the other way, we can keep pattern matching with Haar Cascade to detect nose, but we can add more side nose pictures to detect side face, and improve the accuracy of side face mask-wearing detection.
Furthermore, we can integrate our detection app to a moving robot (like the figure below), and make it patrol in our campus, to notice people wearing their masks on or wearing their masks properly.
Figure 12 [7]
What We Learned from the Project
We learned how to apply our computer vision knowledge to make a useful APP.
Also, from the lecture, we knew there exist some deep learning techniques which we can apply to our project, to use MobileNetV2 to train our model and let the model be used to detect mask-wearing.