Highlights: Face recognition represents an active area of research for more than 3 decades. This paper, FaceNet, published in 2015, introduced a lot of novelties and significantly improved the performance of face recognition, verification, and clustering tasks. Here, we explore this interesting framework that become popular for introducing 1) 128-dimensional face embedding vector and 2) triplet loss function. In addition to the theoretical background, we give an outline of how this network can be implemented…
Read more

Highlights: In this post, we will talk about the importance of visualization and understanding of what our Convolutional Network sees and understands. In the end, we will write code for visualizing different layers and what are the key points or places that the Neural Network uses for prediction. Tutorial Overview: History Introduction Visualization with a Deconvnet Saliency and Occlusion  Visualization and understanding CNNs in PyTorch 1. History A bit of history about CNN’s, back in…
Read more

Highlights: The year 2017 was very fruitful for Google researchers working on semantic segmentation. Their proposed model called the DeepLab was significantly improved over several iterations. In their 4th paper, they present Version 3+ of the same model. In this blog post, we will study the theoretical novelties of this version that utilizes the model developed and popularized in Version 2. We will also see how they frame the first model as the decoder part…
Read more

Highlights: Semantic segmentation is an important subject in Computer Vision that enables a model to label specific regions of an image according to what’s being shown. DeepLab is a state-of-the-art model by Google with many versions making a family of algorithms used for semantic segmentation. In this tutorial post, we will introduce the DeepLab algorithm and specifically talk about the DeepLab v2 that introduced three famous advancements in the field of semantic segmentation. Towards the…
Read more