Hackathon- computer vision

Welcome to our Hackathon-2024. Through this hackathon we are trying to select interns have passion to work in computer vision and deep learning. You can select either C++/Python to solve these coding assignments. Best of luck everyone

• Use issues to ask any queries

• Only make use you use the images given in the dataset rather than using any other public images. That is if a problem has limited dataset that is a constraint and you have to develop a algorithm optimal with respect to the constraint

Problem 1.

Write an algorithm that will detect the defect in the images given below. Breakup of the mark is as given below ** Expecting a general algorithm that works for different diameter, transaltions during image acquisiton etc .

• Flowchart - 20 marks

• Basic algorithm to find defects - 40 marks

• Localizing the defect - 20

• Classify the defect to flashes and cut marks - 20

Good Image

Defective Images

1. Flashes

2. Cuts

• dataset

Problem:2

In manufacturing companies the production lines are optimised for creating good samples while defective samples are very hard to generate and will happen only once in a while. To adress this issue we have to develop quality inspection system from golden samples. In this problem we have given a train dataset of a screw where good samples conists of 2 screws of different length and 4 nuts. Design a anomaly model which can detect both logical and structural changes in the data

• dataset

Problem 3

A footwear company requires to detect the size mismatch of the their pair of footwear(that is sometimes the operator will place two different sizes as a pair ). Select a method to solve this poroblem. The dataset. You are free to chose any model to solve the problem and we had given a mask data. you can use the same for validation or even training depend upon the how you want to solve

Email Your solutions in a github repo to : hr@dhvaniai.com