Description:

Images are inherent part of most of the apps out there. One should be able to identify what's in your images and use intelligence to build products out of it. Building such systems are hard both in terms of R & D effort and building a scalable system out of it. With deep learning it has become far easier and efficient though.
In this proposal i would like to talk about how to build a visual classification system with deep learning involving training and classification engine on the cloud. Best practices with deep learning and transfer learning to boost your accuracy.

I use python, Nvidia based server on AWS.

Below examples are snapshot that i used to build a image tagging engine to classifiy images based on different attributes like clothing type, sleeve type, collar type, pattern and color. Built by scrapping images from different e commerce portal like flipkart, myntra, amazon etc.

<img src="http://bipulja.in/static/images/demo2.png" width="500px"><br> <br> <img src="http://bipulja.in/static/images/demo1.png" width="500px"><br>

Prerequisites:

Python, Machine Learning, AWS

Content URLs:

Home Page : http://bipulja.in<br> GitHub : https://github.com/bipul21<br>

Speaker Info:

ex-flipkart, pythoniast, hacker and a explorer. Now working purely with python with CredR in Mumbai.

Speaker Links:

Quora: https://www.quora.com/profile/Bipul-Jain <br> Twitter : https://twitter.com/_bipul<br> Blog Post: http://bipulja.in/blog/visual-classification-engine-neural-network-5