/Dog-Cat-Image-Classifier

Dog and cat image classifier using feature extraction techniques and machine learning based classification models.

Primary LanguageJupyter Notebook

DOG & CAT CLASSIFIER

Dog and cat image classifier using feature extraction techniques and machine learning based classification models.

alt text

Used feature extractors:

Extractor 1 :

This method of feature extraction consists of:

  1. split the image into NxN sized cells
  2. Assign each neighbouring pixel a value of 0 or 1 depending on the intensity.
  3. Using the neighbour values to make a binary number and assigning that colour to the central pixel.
  4. Once you have done that with everyone calculate a histogram per cell.

Results:

Dataset Accuracy Run Time
"cat_dog_100" 95% 16.30 seconds
"cat_dog_500" 92.5% 4 minutes 26.28 seconds

Extractor 2 :

This feature extraction method involves three steps:

  1. Apply convolution filters to calculate the magnitude and orientation of the gradient.
  2. Divide the image into disjoint blocks and create histograms for each block.
  3. Normalise the histograms.

Results:

Dataset Accuracy Run Time
"cat_dog_100" 92.5% 16.46 seconds
"cat_dog_500" 90% 1 minutes 27.68 seconds

Classifier used:

Scikit-learn SVM

https://scikit-learn.org/stable/modules/svm.html

Explanatory videos (In Spanish):

Video

VIDEO

Code Video

VIDEO

Created by Odei H. and Endika A. 2022-2023