Hands on Advanced Bag-of-Words Models for Visual Recognition. The content of this tutorial is organized around a collection of MATLAB hands-on lab exercises introducing fundamental concepts in visual recognition.
This tutorial has been held in conjunction with ICPR 2014 and ICIAP 2013. It is also regularly held at University of Padova and University of Florence (Italy), in both M.Sc. and Ph.D. programs in computer science and computer engineering.
More info at: https://sites.google.com/site/handsonbow/
Instructors:
- Lamberto Ballan, Ph.D. - University of Padova, Italy (www.lambertoballan.net)
- Lorenzo Seidenari, Ph.D. - University of Florence, Italy (www.micc.unifi.it/seidenari)
The tutorial is based on the Caltech-101 dataset. We provide two subsets for experiments: 4_ObjectCategories and 15_ObjectCategories.
We provide these datasets in two versions: image only and image + pre-computed visual features. You can download the zip files from these urls:
- https://www.micc.unifi.it/downloads/tutorial_bow/images/4_ObjectCategories.zip (33MB)
- https://www.micc.unifi.it/downloads/tutorial_bow/images/15_ObjectCategories.zip (32MB)
- https://www.micc.unifi.it/downloads/tutorial_bow/images_and_feat/4_ObjectCategories.zip (1.2GB)
- https://www.micc.unifi.it/downloads/tutorial_bow/images_and_feat/15_ObjectCategories.zip (1.6GB)
Once you have downloaded the files the default directory in which they should be unzipped is "~/handsonbow/img".
The starting point of the tutorial is the Matlab script exercises.m. In this script you will find several exercises to be completed. For each of them, the instructions/suggestions are provided as comments in the script and you have to add some lines of code where requested.
We provide also a complete version of the same script (exercises_solutions.m).
The file with solutions is intended to be used as a reference and we usually do not provide them in class.