dariahem / dog-classification Goto Github PK

The goal of this project is to make program that can classify dog breed using Convolutional Neural Networks (CNN).

1. The classifier decides if the photo represents the dog, human or category named "others". If the decision is: dog, the classifier should classify the dog breed. If the decision is: human, the classifier should point to which dog breed the human on the photo is similar to.

1. I created the human detection program. For this purpose I used cv2.CascadeClassifier in cv2 libarary.

2. I created the dog detection program and used transfer learning of VGG16 and used all the parameters already trained.

3. I created the classification program to detect 1 of 133 dog breeds (ImageNet data contains 133 sorts of dogs). For this purpose I used transfer learning by VGG16 but changed last classifier linear model to 133 final classes

4. Connected all the steps to make one executable program.