What Deep Machine Learning models can I use for Pre-processing and Feature Extraction of partially occluded faces (nose masks) of individuals in Night Mode.
Article Deep learning techniques for detecting and recognizing face ...
Article Deep representation for partially occluded face verification
https://ieeexplore.ieee.org/document/8832330
https://www.intechopen.com/chapters/79240
https://arxiv.org/pdf/2103.10492.pdf
Depending on the masks included in your dataset and the dataset size, you may consider annotating some data or using conventional image processing methods to generate weakly labeled dataset whereby the label or region of interest is the "nose mask" or "face mask". Then, you can train a deep learning model for mask segmentation/detection. Subsequently, features can be easily extracted from the region of interest generated using the trained segmentation/detection model.
All the best!
Patrice M.
Hi
it depends on a face mask. in some cases, the Dlib library can be useful the position of some key points such as eyes, nose, and lips .... can be identified by utilizing this. also by employing compression methods like the HOG/Haar method of FF2 you can compress the image.
https://iopscience.iop.org/article/10.1088/1742-6596/2181/1/012004/pdf
https://www.frontiersin.org/articles/10.3389/fpubh.2022.955332/full
I hope these two articles are useful for you.
Use a convolutional neural network from scratch if you have a lot of images, or use transfer learning methods if you have a small number of images