End to End Computer Vision paradigm with respect to advanced deep learning techniques.

Deep learning based approaches to solve image classification have become a core technology in AI, largely due to developments in computing powers and digital data. However image classification gained popularity beyond academic circle with the advent of visual object recognition challenge.

In this talk, we will walk through the journey of deep learning in the field of computer vision. The main focus will be on the most recent and advanced technique for image classification and object detection .We will walk through various classical architecture and in the journey will learn concepts like padding, max pooling .

To make the talk more interactive we will show live demo and code run of various use cases like car detection for autonomous driving.

Keywords: Object detection, Transfer Learning, Art transfer, Max-Pooling,Padding


Outline/Structure of the Tutorial

  • What is CNN ?
  • Classic Architecture
    • Resnet
    • Inception
    • Others
  • Data Augmentation
  • Transfer Learning
  • Fine Tuning
  • Case Study and applications
    • Autonomous Driving Car Detection
    • Overview of transfer learning

We will explain implementation of case study with jupyter notebook to get hands on experience.

Learning Outcome

By the end of this session, the audience will have a clear idea about how to tackle various problems of Computer Vision. The audience will get fundamentals of how various deep learning models work and the latest innovations that have taken place in Computer Vision.

Target Audience

Anyone who is interested in latest developments in deep learning and wants to understand how modern techniques have evolved in the field of computer vision. 

Prerequisites for Attendees

Basic understanding of deep learning and how neural networks are trained. Beginner level knowledge about Python and Keras will be helpful in understanding the concepts more efficiently.

schedule Submitted 4 months ago

Public Feedback

comment Suggest improvements to the Speaker
  • Anoop Kulkarni
    By Anoop Kulkarni  ~  4 months ago
    reply Reply

    Thanks for your proposal. The current reading of the proposal looks broad based to me. Rather than making it a generic talk, is it possible to use just one of the  use-cases (you have mentioned 3 which may be too many) and run through challenges that one faces designing a DL model for it?

    Do share your thoughts, in case I am missing something.



    • Pushkar Pushp
      By Pushkar Pushp  ~  4 months ago
      reply Reply

      Thanks for your valuable feedback.

      Our main purpose for this talk was to ensure that user would be able to build ,train ,tune a DL model for various computer vision aspects.The crux was most talk on DL/ML somehow misses the mathematical aspects of model building . So we decided to even cover the mathematics/statistics behind all these stuffs. 

      The Case Study which we will Cover in detail will be one of the applications in autonomous driving. Technically speaking it will broadly cover the below mentioned concepts.

      • Object Detection Algorithms such as Faster R-CNN, Resnet
      • YOLO in depth explanation and implementation
      • Understanding Intersection over Union and why is it used as metric in object detection.
      • What is Non-max Suppression
      • Lastly we will walk through Tensorflow implementation of some other detection models. 
      • Analysis of loss functions ,hyper-parameter tuning.

      Case study on Art Transfer is more of fun activity to engross audience or else they can takeaway this notebook and play around it to grasp CNN concepts.

      We will give enough input on transfer learning to guide the audience for image classification problems.

       In short the focus we will be to cover two main used cases in real life:

      • Object Detection
      • Classification