This is my project during Mitacs internship at the university of Saskatchewan.

Introduction

The program automatically detects cells in the ALL-IDB dataset.

Proposed Method

Counting problem

Classification

Requirements

1. numpy >= 1.9
2. opencv >= 2.4.10
3. skimage >= 0.12
4. matplotlib >= 1.4.3
5. pymorph >= 0.96
6. spams (install from source)

Programs

Counting Framework

Usage

Run python program_nolearing.py -h for more details about using the program.

Examples

python program_nolearing.py -d training.txt -v: count all images which are located in file training.txt and visualize the result.

python program_nolearing.py -d training.txt -o result.txt: count all images which are located in file training.txt and write the result to file result.txt

To load data from training.txt or problem.txt, you must download the dataset and copy all images and xyc files into folder data. Because of the copyright of the dataset, I can not make a copy version in this repository.

Learning Framework

Usage

Run python program_fusion.py <training path> <testing-path>

Examples

1. python program_fusion.py train/allidb_1.txt test/allidb_1.txt

ALL IDB 2

Because the framework is in progress as well as this version is not the final one. To evaluate the performace of the ALL-IDB 2 dataset, you have to use these scripts which have the "allidb2_" prefix.

For example, to run the fusion of classifiers framework, type this script to your console terminal:

python allidb2_fusion train/allidb2_1.txt test/allidb2_1.txt or

python allidb2_fusion it uses file train/allidb2_1.txt as training samples and file test/allidb2_1.txt as testing samples.

Development

[Updating]