Semantic Segmentation in Images using Pytorch

Semantic Segmentation What is Semantic Segmentation? Semantic Segmentation is identifying every single pixel in an image and assign it to its class . As displayed in above image, all pixels of an object are assigned same color and it is done for all the objects. Note : It doesn't tells us about different instances of …

Transfer Learning – Using Inception V3 for developing Image Classifier (Part 2)

Recap - In part 1 , I briefly explained about transfer learning and how to use InceptionV3 trained model to generate feature vector(aka feature map) for an image. In this post, I will describe how to train new image classifier using the feature map generated by InceptionV3 trained model. What? Generate Feature map for images …

Object Detection – R-CNNs explained

What? R-CNNs is a Region based Convolutional Neural Networks and was state-of-art architecture for object detection in computer vision at the time of its formulation. Objects detected, Image by tensorflow How? Region ProposalExtract around 2k regions from an image using Selective SearchFeature GenerationCreate feature vector for each region proposed using CNNs trained for image classification, …

Transfer Learning Explained – Using Inception V3 for developing Image Classifier (Part 1)

Transfer Learning is a paradigm used in machine learning which involves using knowledge gained from previous solved task to solve and improve learning of new related problems. Why? Related problems exhibits commons patterns and features which are used during learning to solve these problems.Recently , Deep Learning has been in use and it is outperforming other previous methods …

CNNs using tensorflow

This tutorial briefly talk about the implementation of Convolutional Neural Networks using tensorflow's keras library. I have developed an Image Classifier using CNNs for classifying mnsit fashion dataset. Output of training -x_train shape: (60000, 28, 28) y_train shape: (60000,) x_train shape: (55000, 28, 28, 1) y_train shape: (55000, 10) 55000 train set 5000 validation set 10000 test …

Autonomous Query Clustering using DBSCAN

In this post, I am going to talk about Un-supervised Clustering of Syntactically similar queries in python. There are multiple algorithm for clustering with k-means being the popular one but the disadvantage with k-means is that you have to provide k, as number of clusters you want to create. Here , I have used DBSCAN …

Maximum Matching(BPM using DFS)

Maximum Bipartite Matching in a graph This one is easy to implement and for smaller graph. It uses DFS to make cut and establish matching to produce maximum matching. Thsi DFS takes adjacency matrix as input and update the matchL[] and matchR[] which store the node that matched to the node in Left and the …

MayTheBestPetWin : Div 2 1000

Problem link:- http://community.topcoder.com/stat?c=problem_statment&pm=12779&rd=15705 Editorial :- http://apps.topcoder.com/wiki/display/tc/SRM+593 Reduced to knapsack type problem. My solution in c++:-

Maximum Xor Secondary

Link :- http://codeforces.com/contest/281/problem/D Level :- Div:-2, D. Problem :- Given an array consisting of N integers.For every l<r such that 0<=l and r<N-1 find out the maximum of Xor of maximum and second maximum number of sub-sequence array[l],......array[r]. Solution :- Segment tree + recursion + memoization. Let in range [a,b], {i,j} be the position of maximum(max1) …

RELINETS spoj

Problem :- Relinet Problem link:- https://www.spoj.com/problems/RELINETS/ Algorithm used :-Tarjan's algortihm This problem is bi-connected graph and need to find that Bi-connected graph which have minimum edges-cost-sum. As n and m are small ,so I use binary representation method to choose subset of edges from the  set of edges and then check for Bi-connectivity in the …