sister Vector Vector Vector

GitHub Repo https://github.com/Ajay2521/Face-Recognition-using-Siamese-Network

Ajay2521/Face-Recognition-using-Siamese-Network

Deep learning algorithms fail to work well if we have only one training example. One-shot learning is a classification or object categorization task in which one or a few examples are used to classify many new examples. The principle behind one-shot learning is Humans learn new concepts with very little supervision. ### Problem in COnvNet - A small training set is really not enough to train a robust neural network for this task. The feature vectors trained do not contain important information that can be used for the recognition of future images. - Retraining the ConvNet every time the number of classes or dataset is increased is way too time and resource consuming. ### Solutions to one-shot learning #### 1. Siamese network for one-shot learning ![image-2.png](attachment:image-2.png) #### Siamese networks are based on a similarity function, which does not require extensive training #### Takes input two images—one being the actual image and other the candidate image—and outputs the similarity between the two. The two input images, if very similar, output = lower value The two input images, if not similar, output = Higher value degree of difference between the two images is compared with the threshold value(𝜏) (which is a hyperparameter), if degree of difference between 2 image < threshold -> output = same person if degree of difference between 2 image > threshold -> output = different person #### Both the images to be compared are passed through the same networks, called sister networks, having the same parameters and shared weights. images are passed through a sequence of convolutional, pooling, and fully connected layers, and end up with a feature vector of a fixed size, say, 128 denoted by f(x1) as an encoding of the input image x1.

GitHub Repo https://github.com/sam-lb/VectorGraph2D

sam-lb/VectorGraph2D

2D Vector field calculator and visualizer. Sister software to MathGraph3D.

GitHub Repo https://github.com/tcho6319/Cosine-Similarity-of-Dickens-and-Brontes

tcho6319/Cosine-Similarity-of-Dickens-and-Brontes

This Jupyter Notebook file was created with Ian Wilke-Tomasik to extract the words and word counts of the Bronte sisters' novels and two Charles Dickens novels. The word-count vectors for each novel were used to compute cosine similarities.