Deep Eye - Automatic Product Categorization for E-commerce using Tensorflow
E-commerce websites are remarkably complex systems with heavy data-intensive and large-scale problems that need solving for a better customer experience.
This talk will focus on the problem of categorization, specifically image categorization at scale in near real-time performed on semi-structured and unstructured image sets. We’ll look at a cost affective way of using GPU instances to scale up product categorization for large scale datasets using Tensorflow, an open source machine learning library open sourced by Google’s Machine Learning Team.
- Introducing large scale image datasets, around 10 Million
- Overview of highly parallel neural net modelling and training with TensorFlow
- Cost effective GPGPU with Tensorflow
- Accuracy and scaling data models in near real-time
- Interpreting training results and validating models
- Storing trained data sets for efficiently categorizing new Images real-time
Enthusiasm to work on large scale image processing and experience with any machine learning models.
Aditya Prasad is a Lead Data Scientist at Craftsvilla.com, where he builds and manages the data pipeline and intelligence to enable India’s biggest online ethnic store.
Earlier, he was a Data Scientist at Housing.com where he oversaw and led multiple efforts that enabled higher conversions and scaled the business. He has previously given talks on scalable data pipelines and building impactful models for e-commerce websites.
He is a B.Tech from IIT Bombay with 5 years of development experience in Python and is often involved in meet-ups and active in the technical community.