• face verification
• convolutional neural networks
• deep learning
• low resolution images

The goal of this thesis to investigate how low resolution images affect the accuracy of deep convolutional neural networks (CNNs) in face verification which is the task of determining if two face images belong to the same person. We manipulated photos from the Labelled Faces in the Wild dataset in order to obtain different levels of images resolution used to test state of the art CNNs. We considered the comparison of couple of images at same resolution dataset and also the case in which one of the two images belongs to a high resolution dataset, while the other one to a low resolution dataset. The results show a decrease in performances of state-of-the-art CNNs trained on the high resolution images. In order to achieve better accuracy on low-resolution images and on mixed resolution, we tested the siamese learning approach for learning an embedded layer using a contrastive loss. In particular, we fine-tuned a pre-trained CNN using the siamese approach on high resolution images and then we selected the best neural network for training on a dataset containing both high and low resolution images. The framework used for training and testing phases is Caffe that provides all the layers for deep learning; we implement a custom layer in python to manage the special input of the siamese network. In order to speed up the entire process, all the training and testing phases are executed on two NVIDIA's GeForce GTX 1080 with 8 GB frame buffer each.
The results show that a low resolution training can actually improve performances. In particular we achieved good results when the test set was composed only of low resolution images and the best results when the test set included both high and low resolution images, so the presence of high resolution images is however preferable in order to achieve such outcomes. Correspondingly to this increment we obtained a decrease in performances of test set composed only of high resolution images. Such decrease is however acceptable and reasonable considering the kind of training we have executed.