Real-time open set facial recognition system
Despite the world nature is dynamic and open, most of the recognition systems presume that the world is static and using a closed world model which treat every object must be known class in prior. However, in face recognition, we do not have the entire set of all possible faces. This project formula...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Year Project Report |
| Language: | English English |
| Published: |
Universiti Malaysia Sarawak, (UNIMAS)
2017
|
| Subjects: | |
| Online Access: | http://ir.unimas.my/id/eprint/20958/1/Real-time%20open%20set%20facial%20recognitions%20system%20%2824%20pgs%29.pdf http://ir.unimas.my/id/eprint/20958/8/Lim%20Yoong%20Kang%20ft.pdf http://ir.unimas.my/id/eprint/20958/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Despite the world nature is dynamic and open, most of the recognition systems presume that the world is static and using a closed world model which treat every object must be known class in prior. However, in face recognition, we do not have the entire set of all possible faces. This project formularizes a generalized method for real-time recognition of open set recognition approach. The proposed approach enables the model to recognize an infinite set of faces in a myriad of unknown faces and unknown, unseen, new or novel faces. The capability of constantly recognize unknown in real-time are the highlight of this system as an unknown face are treated as a valid outcome. This system utilized Haar Cascade to detect faces, PCA and LDA to extract the most significant data to describe the face, and Approximate Nearest Neighbour to recognized the previous known face and identify unknown faces via distance metric. Once unknown faces are identified, it will be learned and be labeled. Hence when the faces appear the second time, it will be recognized. In result, 60% accuracy is archived after fine tune the hyperparameter of feature selector and the distance metric threshold. |
|---|