+1 -1 +62
Vote on this proposal

Computer Vision using Python: Can webcams serve as eyes for the blind?

by Sanyam Garg (speaking)

Section
Core Python
Technical level
Intermediate

Objective

The talk will help introduce the audience with the recent developments in computer vision. I will be covering the current trends in object recognition and scene understanding. Finally, I will explain how we can code to detect humans in a live video.

Description

The best results for object detection in images were obtained when a new descriptor using grids of histogram of oriented gradients was suggested by Dalal and Triggs in their paper: "Histogram of Oriented Gradients for Human Detection". The descriptor outperformed previous feature sets for object/human detection. In my talk I will be explaining the following concepts:

a) How HOG descriptor is formed for every image
b) Training the SVM classifier using the pedestrian dataset
c) Finding objects/humans in a new image
d) Utilizing the above concepts to find humans in a live video.

The entire concept will be implemented using the OpenCV library in Python.

Requirements

  1. A laptop with Windows OS installed (preferable) . You may work on any OS but I will be using Windows.

  2. Python 2.7. Download Link: https://www.python.org/download/

  3. OpenCV for Python
    Downlaod link: http://opencv.org/downloads.html
    Installation guide: http://docs.opencv.org/trunk/doc/py_tutorials/py_setup/py_setup_in_windows/py_setup_in_windows.html

  4. An external webcam. (the front camera on the laptop will also work)

Speaker bio

I am a student at BITS-Pilani Hyderabad Campus with more than 2 years of experience in Python and Matlab. I am a computer vision enthusiast with significant experience in active contours (Snakuscules) and color constancy algorithms (Intern - IISC, Bangalore). One of my research works on estimating glucose concentration using image processing is currently in the proceedings of 'IEEE International Conference on Image Processing'. I have lately with working on the concepts of machine learning that can be incorporated in vision systems.

Comments


  • 1

    [-] Ashwin Ashok 292 days ago

    very nice...
    very interesting. really looking forward to it


  • 1

    [-] Utkarsh Sinha 270 days ago

    Is this a workshop?


    • 1

      [-] Sanyam Garg 264 days ago

      No this is a simple talk wherein I will be explaining the concepts behind detecting humans/pedestrians in an image. The code may be explained depending upon the availability of time.

Login with Twitter or Google to leave a comment →