Improving Recognition Performance for Low-Resolution Images Using DBPN
| Autor: | Syed Ariff Syed Hesham, Lijun Jiang, Keng Pang Lim |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Neutral network
Landmark Computer science business.industry Low resolution ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Python (programming language) Facial recognition system Expression (mathematics) Image (mathematics) Computer vision Artificial intelligence business Image resolution computer computer.programming_language |
| Zdroj: | 2021 IEEE 16th Conference on Industrial Electronics and Applications (ICIEA). |
| DOI: | 10.1109/iciea51954.2021.9516374 |
| Popis: | Artificial Intelligence (AI) has evolved and been used for many different purposes. Python is one of a well-known language when it comes to AI due to the presence of the libraries like TensorFlow, Keras and Pytorch. The Aim of the project is to recognize faces from low resolution (LR) images, small size or poor-quality images with varying pose, illumination, expression, etc. This issue has received much attention due to the increasing demands for long distance surveillance applications. In this paper, an overview on the problems and the expected progress are presented. A solution based on Deep Back-Propagation Network (DBPN) is proposed to improve accuracy of the face recognition by using super resolution(SR) approach, which include detection / plotting of the facial-features/landmarks in an image, then using this landmark to align the image, training on a state-of-art Neutral Network for SR of LR image, etc. Experimental results shows that the proposed system achieved8.41 % precision and 8.38 % recall rate for a 4425 persons dataset. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|