Dog Breed Classifier

Name: Dog Breed Classifier

Author: Sharome Burton

Date: 07/20/2021

Description: Machine learning model used to determine the breed of a dog from a given image.

Kaggle: https://www.kaggle.com/sharomeethan/dog-breed-classifier

Colab: https://colab.research.google.com/drive/1wLBBuwKx4a9w3jTectO0nzCzU9d3w_BB?usp=sharing

1. Problem definition

How well can we identify the breed of a dog from a given image?

2. Data

We are provided with a training set and a test set of images of dogs. Each image has a filename that is its unique id. The dataset comprises 120 breeds of dogs.

• train.zip - the training set, we are provided the breed for these dogs
• test.zip - the test set, we must predict the probability of each breed for each image
• sample_submission.csv - a sample submission file in the correct format
• labels.csv - the breeds for the images in the train set

There are 10,000+ labeled images in each set.

source: https://www.kaggle.com/c/dog-breed-identification/data

3. Features

• id - a unique identifier for each image
• breed - the breed of the dog, eg. * affenpinscher * afghan_hound * african_hunting_dog * airedale * american_staffordshire_terrier

4. Evaluation

Goal: Determine the breed of a dog in a given image with >75% accuracy.

The evaluation is a file with prediction probabilities for each dog breed of each test image. Submissions are evaluated on Multi Class Log Loss between the predicted probability and the observed target.

source: https://www.kaggle.com/c/dog-breed-identification/overview/evaluation