Low-resolution Face Recognition & Identification using Super-Resolution

• A face recognition & identification program intended to detect & identify faces from pictures that are of size 30x30 pixels using EDSR image super-resolution.

• Achieved accuracy of at least 90% on samples across 30 distinct faces.

• All faces have been taken from the LFW3D face samples, real-life faces are used as well

  • Take note that some personal info (especially the real-life faces) have been redacted from the samples & documentation.

Technologies Used

• Python
• OpenCV
• EDSR (Enhanced Deep Residual Networks) image super-resolution
• Bicubic & nearest neighbor scaling

Installation

• Download the repository and unzip it
• Install Anaconda Python IDE
• Install relevant libraries
• Select the version to run (for KNN model, use the knn-version; for SVM model, use the svm-version)
• Open the relevant .ipynb notebook in the folder in sequence

Credits

• LFW3D for the faces supplied
• Mr. Geitgey for his high-resolution face recognition & identification trained model.