[Pycon] [new paper] "Chirag Shah" - pyCAIR: Content Aware Image Resizing in Python

[info a pycon.it]

Title: pyCAIR: Content Aware Image Resizing in Python
Duration: 45 (includes Q&A)
Q&A Session: 15
Language: en
Type: Talk

Abstract: **What if we had our own tool that can manipulate an image(i.e. crop/resize) on the basis of content within the image?**  
In this talk, I will speak about a simple yet very powerful **image manipulation mechanism.**  

The naive user utilizes the services of any standard toolkit, be it a web service or a remote application for image manipulation. The black box approach to this process is: A user provides an image and other parameters as input to the toolkit which in turn produces the results and returns it back to the user. Often these results are not up to the mark. The image sometimes gets distorted, misaligned or blurred. These standard mechanisms aren't the best way to manipulate images.  

Deviating from the standard mechanisms, I would like to talk about a technique called as **Content aware image resizing**. The primary factor in this technique is **the content**. It is the content which drives the entire technique. The image is cropped, enlarged or modified keeping in mind the primary factor.   

I will talk about an algorithm called as **Seam Carving** which is used under the hood to achieve the aforementioned technique. **It is this algorithm and the power of Python libraries**, that makes this technique perform better than the standard mechanisms.  

----------

**Outcomes:**

 - To utilize pyCAIR and use it for image processing.
 - Build projects which can be efficiently utilized in real life.
 - Understanding energy concepts and basic of Dynamic Programming
 - Understand why documentation is necessary and its the only judge of your project.

----------

**Agenda of Talk:**

1. Introduction(0-7 min): Basics of seam carving, how seam carving algorithm works under the hood.  
2. Basics revisited(8-16 min): Understanding energy concepts pertaining to Images, basics of computer vision and dynamic programming.  
3. pyCAIR in a nutshell(16-22 min): Walk over the pseudo code of algorithm and its implementation in Python.  
4. Comparative analysis(22-25 min): pyCAIR vs other mechanisms.  
5. Q&A Session(25-30 min)  

Tags: [u'graphics', u'Python', u'image-processing', u'deep learning', u'documentation', u'scientific-computing']