1 / 13

CVAC: Computer Vision Algorithm Collection Review

CVAC: Computer Vision Algorithm Collection Review. Thomas McGrath CS-6963 Summer 2013. What is CVAC? (https://github.com/NPSVisionLab/CVAC/wiki/Downloading-CVAC). Developed by Professor Mathias Kolsch and his team at the Naval Postgraduate School, Monterey, CA.

wynn
Download Presentation

CVAC: Computer Vision Algorithm Collection Review

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. CVAC: Computer Vision Algorithm Collection Review Thomas McGrath CS-6963 Summer 2013

  2. What is CVAC? (https://github.com/NPSVisionLab/CVAC/wiki/Downloading-CVAC) • Developed by Professor Mathias Kolsch and his team at the Naval Postgraduate School, Monterey, CA. • Main goals– to make usage easier for those not desiring to delve deeper into the metadata and algorithms within the computer vision field and to use remote datasets generated by the computer vision community (i.eLabelMe) • Python-based modules that allow for analyzing a group of images by various algorithms that focus mainly on category recognition. • Utilizes open source software to manage remote connectivity (ICE) and support computer vision processing (OpenCV).

  3. Terminology • Algorithm – specific instructions on how to process a function (Rogers, Jr, Hartley (1987). Theory of Recursive Functions and Effective Computability. The MIT Press. ISBN0-262-68052-1 (pbk.)) • Corpus – the metadata for a collection of images or videos • Detector - to rapidly find likely regions where particular objects might exist. • Ground truth – known good data (usually well labeled image databases or quantitative data to train a detector) • Training – using specifically chosen images to train a detector to correctly assign labels to categories of items • Testing – utilizing a detector to label categories of items within an image

  4. Process • Feature Detection – low level pixel analysis of image • Classified by what they look at: • Edges • Corners (refer to point-like features in an image) • Blobs (regions of interest or interest points) • Algorithm will look at detected features • Note: This is where most of the processing time is consumed.

  5. Algorithms • Bag of Words • computes the distribution (histogram) of visual words found in the query image and compares this distribution to those found in the training images. • Viola-Jones Object Detection Framework • the first object detection framework to provide competitive object detection rates in real-time proposed in 2001 by Paul Viola and Michael Jones. (Viola, Jones: Robust Real-time Object Detection, IJCV 2001) • Many Others

  6. Operating Environment • Windows 7, 64-bit • Python 2.6 (32-bit) • CVAC binary, version for Windows • Start “services.bat” file • Ensure it is configured for the following: • CVAC installation directory • Python 2.6 execution path (to include executable file “python.exe”) • Java runtime executable path (to include java executable “ • Download and install PIL for windows (http://www.pythonware.com/products/pil/) • Download and install Matlab Runtime (R2012b, 32-bit http://www.mathworks.com/products/compiler/mcr/ ) • Download and install zlib (http://www.zlib.net/)

  7. Process • Feature Detection – low level pixel analysis of image • Classified by what they look at: • Edges • Corners (refer to point-like features in an image) • Blobs (regions of interest or interest points) • Algorithm will look at detected features • Note: This is where most of the processing time is consumed.

  8. Algorithms • Bag of Words • computes the distribution (histogram) of visual words found in the query image and compares this distribution to those found in the training images. • Viola-Jones Object Detection Framework • the first object detection framework to provide competitive object detection rates in real-time proposed in 2001 by Paul Viola and Michael Jones. (Viola, Jones: Robust Real-time Object Detection, IJCV 2001)

  9. Operating Environment • Windows 7, 64-bit • Python 2.6 (32-bit) • CVAC binary, version for Windows • Start “services.bat” file • Ensure it is configured for the following: • CVAC installation directory • Python 2.6 execution path (to include executable file “python.exe”) • Java runtime executable path (to include java executable “ • Download and install PIL for windows (http://www.pythonware.com/products/pil/) • Download and install Matlab Runtime (R2012b, 32-bit http://www.mathworks.com/products/compiler/mcr/ ) • Download and install zlib (http://www.zlib.net/)

  10. Demonstration

  11. Challenges / Tips • PythonPath • Use the run.bat file so PythonPath is set just prior to running scripts. • Run the program from the root / CVAC installation directory. • Run the “prerequisites.py” script first to ensure your system is configured properly. • Ensure that the network you are operating on allows for calls to remote server ports. • If access is blocked you will most likely see the error “Ice.ConnectionRefusedException: Unknown error”

  12. References / Additional Information • Documentation for the Easy Module http://movesinstitute.org/~kolsch/CVAC/Easy.pdf • Online documentation for CVAC http://www.movesinstitute.org/~kolsch/CVAC/html/index.html • LabelMe Information (http://labelme.csail.mit.edu/Release3.0/) • Excellent reference for Computer Vision by Dr. Richard Szeliski. http://szeliski.org/Book/ • Computer Vision Sample Data http://www.cvpapers.com/datasets.html • Cal Tech Image Database http://www.vision.caltech.edu/Image_Datasets/Caltech101/

  13. Additional Information • Third Party Support to CVAC: • ICE - Internet Communications Engine - handles the connections "behind the scenes" • http://www.zeroc.com/ice.html • OpenCV - OpenCV (Open Source Computer Vision Library) is an open source computer vision and machine learning software library. • http://en.wikipedia.org/wiki/OpenCV • One of the best collections of Computer Vision information I reviewed. • http://www.computervisiononline.com/ • Research on More Training Data versus Better Models • http://web.mit.edu/vondrick/largetrain.pdf

More Related