IMAGE CLASSIFICATION USING COMPUTER VISION

Live projects are about bringing professional management experience to students which help them in the beginning of their career as well as in the long-run. IFIM Business School has bring out this opportunity to their students by collaborating with renowned institutions all over the global. Insofe, is one of the institution that have team up with IFIM and gave the best out of it to the students.

 

 

The live project main objective is to classify the product images especially grocery products through image recognition by using Convolutional Neural Network (CNN). This mainly depicts into taking an input image, process it and classify it under given certain categories. The data consists of 5000 images belongs to certain classes or categories with set target accuracy of 75% accuracy. The data is loaded using the image data generator followed by pandas data frames. The CNN model is split up into test data and train data with series of convolution layers using kernals. After trying different architectures and parameters at different stages we finally ended up with using InceptionV3 architecture and by using regularizer with a penalizing value of 0.1, optimizer SGD with a learning rate of 0.0001 and momentum of 0.99. The test data accuracy is achieved as of 78% and train data accuracy with 86%. The main limitation of the model is faced during data ingestion to class prediction.

 

This project mainly helps to reduce the employee mundane work to check upon the stock physically rather the employee can simply detect the stock range by using this simple machine learning technique. Besides, this project has given the students an extensive range of knowledge in the area of AI & ML techniques and how it is used in various unknown aspects of everyday life.

 

Koushik Varma
Student at IFIM Business School