Machine learning is being used by people in various ways knowingly or unknowingly. Each time when we Google or Bing search we get the best results because they have used machine learning to rank page. The world biggest search engine giant Google offers recommendation and suggestion based on the previous user searches. In 2012, Google introduced Knowledge Graph -an algorithm used to decipher the semantic content of a search query. Facebook uses a face recognition…
According to Wikipedia “Mathematical morphology (MM) is a theory and technique for the analysis and processing of geometrical structures, based on set theory, lattice theory, topology, and random functions”. It is mainly applied to digital images for image processing. There are many operations of mathematical morphology but mainly used operations are dilation for increasing the image regions, erosion for shrinking the image regions and thinning for structured erosion using image pattern matching. Basic Operations of…
HoGs are a feature descriptor that has been widely and successfully used for object detection. It presents objects as a single feature vector as oppossed to a set of feature vectors where each represents a segment of the image. It’s computed by sliding window detector over an image, where a HOG descriptor is a computed for each position. Like SIFT the scale of the image is adjusted (pyramiding). HOGs are often used with SVM (Support…
Support Vector Machine is a supervised learning algorithm which can be used for both classification or regression problems. In this algorithm, data points are plotted in an n-dimension space (where n is a number of features) with the value of each feature being the value of a particular coordinate. Then, classification is done by finding hyper-plane separating between two classes. Support Vector Machine is a frontier which best segregates the two classes (hyper-plane/ line). Support…
Mask RCNN is Faster RCNN (object detection with bounding boxes) with a mask on it. It is developed by Facebook AI Research (FAIR). The output from Mask RCNN is similar to YOLO (You Only Look Once) but the implementation strategy is different. What does mask do in Mask RCNN? Mask features labels each pixel and compares each pixel with an object. Here is the link of the paper written the Facebook AI Research (FAIR) on…
“YOLO” or “You Only Look Once”, is a massive Convolutional Neural network for object detection and classification. YOLO9000 is high speed, real-time object detection algorithm that can detect on over 9000 object categories. YOLO was originally written in a framework called darknet. Darknet is an open-source deep learning framework written in C and CUDA. It is fast, easy to install, and supports both CPU and GPU computation. If you want to know more follow this…
The process of repeatedly nudging an input of a function by some multiple of the negative gradient is called gradient descent. It’s a way to converge towards some local minimum of a cost function basically valley in a graph. According to Wikipedia, Gradient descent is a first-order iterative optimization algorithm for finding the minimum of a function. To find a local minimum or minimum cost of a function using gradient descent, one takes steps proportional…
In Artificial Neural Network (ANN), the activation function of a neuron defines the output of that neuron given a set of inputs. For Neural Network to achieve maximum predictive power, we must apply activation function in the hidden layers. An activation function allows the model to capture non-linearities. Image 1 below from study.com gives examples of linear function and reduces nonlinear function. If the relationships in the data are not straight line relationships we will…
Transfer Learning is a machine learning technique where model trained on one task could be applied to the different but related problem. For example (from Wikipedia), knowledge gained while learning to recognize cars could apply when trying to recognize trucks. This model is retrained in a similar problem which will drastically cut down the training time. To conduct deep learning could take days, but transfer learning helps to conduct the work in a very short…
Reinforcement learning is based on reinforcement theory in which a combination of reward and/or punishment is used to reinforce the wanted behavior or extinguish unwanted behavior. Software agents interact with the environment. In reinforcement learning agents can automatically figure out how to optimize their behavior given a system of rewards and punishment. It draws inspiration from behavioral psychology. It has applications in many fields. Such as: Economics Genetics Game Playing In 2016, reinforcement learning was…