-
Notifications
You must be signed in to change notification settings - Fork 0
GSOC '14 Proposal
Python Software Foundation,
SciKit-image
- Name: Syed Tabish
- Email: syedtabish.a@gmail.com
- Telephone: +917893498365
- source control username: syedTabish@GitHub
- IM syedtabish.a --- Google Talk
- Blog: tabishgsoc14.blogspot.in
- University: International Institute Of Information Technology,Hyderabad
- Major: Electronics And Communication Engineering + MS
- Current Year: 4th year
- Expected Year of graduation: 2015
- Degree: (Bachelors in Technology, Honors + MS by research)
We propose a python module for visual saliency computations. The library would involve several popular visual saliency models with an easy and complete API for effective use. The library is expected to be an attractive addition to Python considering the fact that, visual saliency has been turning up as an attractive topic in several computer vision conferences and problems.
Visual saliency is a concept inspired by a study of early visual processing in the human brain. Several studies and experiments have indeed shown, that prior to processing the visual information, the brain selects relevant regions in the visual field in the order of their preference for processing. This enables the brain to preferentially process visual information more critical to the survival or completion of some task by the individual. This behavior of the visual system is thought to be directly stemming from the fact that the brain despite being enormously capable of computations, still falls short of resources when it comes to visual processing. This is because visual information is full of long-range and many semantic contexts which do put a lot of stress on the computational resources of the brain, not to forget the fact that this has to be a real-time process. Thus, visual saliency is nature's way of pre-processing with the objective of selecting relevant regions of the image thus giving it a sparse representation and reserving more intensive operations for the relevant regions in the order of their preference. Recent times have seen a proliferation of discussions about visual saliency in computer vision community and wide array of papers have appeared in the context of visual saliency in computer vision areas. However, there is a severe lack of a self-contained library of collection of important saliency models which can readily be used by others. Such a self-contained library coupled with the powerful capabilities of Python can be very helpful for computer vision community. It will also through its availability and the open-source nature of Python encourage the interest of the community in the applications of Visual Saliency.
The deliverables for this project include a library for visual saliency and associated documentation.
The library features include :
a) Availability of relevant bottom-up (task-independent) and top-down (task-dependent) saliency models as ready-made functions with a nice API providing access to arguments which provide for maximum tuning of saliency parameters. The presently planned models include :
- Itti-Koch Model
- GBVS Model
- Spectral Residue (SR) Model
- Extended Spectral Residue (ESR) Model.
- Attention by Information Maximization (AIM) Model.
- Achanta's LAB saliency
- Image Signature Saliency Model.
###Before April 21st
- To familiarize myself with PEP-8 Style guide for python code.
- Already familiar with git.
- Understand the community standards and documentation standards.
- To finalize the API for each saliency model so as to ensure hassle-free development with a well-focused objective.
- To stay in touch with the mentor to discuss and receive feedbacks on API design.
- To do a basic self-coding of the models to gain experience of coding while following a proper community recommended development, testing and version control system.
- To effectively equip the library with at least half of the models and in a fully working state.
- To do extensive testing and evaluation of the library.
- To discuss the contemporary design with the mentor and receiving and making changes as needed.
- Modify the future designs as per the needs.
- To finish the implementation and testing of the library.
- To Prepare the documentation following the community standards.
- To discuss the work and documentation with the mentor and receiving and incorporating the feedbacks received.
2 weeks have been kept for buffer to adjust for unexpected delays or issues.