DSND Capstone Project (Dog Breed Classifier)

1. Project overview

The main focus of project is to build a dog breed classification algorithm using Convolutional Nerual Networks to detect a dog or human in a user provided image and return the following results:

• The predicted dog breed if a dog is detected in the image.
• The resembling dog breed if a human is detected in the image.
• Report an error if neither is detected in the image.

2. Software packages

python (v3.6.3)
glob
random
numpy (v1.12.1)
pandas (v2.23.3)
matplotlib (v2.1.0)
sklearn (v0.19.1)
keras (v2.0.9)
cv2 (v3.3.1)
tqdm (v4.11.2)
PIL (v5.2.0)

3. Files

The files included in this project are:

• dog_app.ipynb: The Jupyter notebook with all code used in this project

• images folder: Example images used in the dog_app.ipynb

• new_images folder: User provided images downloaded from the internet for model assessment

• extract_bottleneck_features.py: The function to extract the bottleneck features

• haarcascades folder: The Haar feature-based cascade classifier for face detection using OpenCV's implementation

• bottleneck_features folder: Bottleneck features from pre-trained model (CURRENTLY UNAVAILABLE DUE TO LARGE FILE SIZE)

• saved_models: All trained models generated in dog_app.ipynb

• README.md

4. Project workflow

Step 0: Import Datasets:

Import dog images and create:

• Numpy arrays containing file paths to train, validation and test images
• Numpy arrays containing onehot-encoded classification labels for training, validation and test
• A list of string-valued dog breed names for translating labels, dog_names

Import human images and create:
A Numpy array of human image file paths, human_files

Step 1: Detect Human

Create and assess a human face detector to detect human faces in images using OpenCV's implementation of Haar feature-based cascade classifiers. One of these detectors have been downloaded and stored in the haarcascades directory.

Step 2: Detect Dogs

• Pre-process the dog images and use a pre-trained ResNet-50 model to detect dogs in the images.
• Create and assess the dog detector.

Step 3: Create a CNN to Classify Dog Breeds (from Scratch)

• Pre-process the dog images
• Build a multi-layer CNN model architecture
• Compile and train the built model
• Load model with best validation loss
• Test the model accuracy (Test accuracy should be greater than 1%)

Step 4: Create a CNN to Classify Dog Breeds (using Transfer Learning)

• Obtain Bottleneck Features
• Build model architecture using a pre-trained ResNet-50 model
• Compile and train the transfer learning model
• Load model with best validation loss
• Test the model accuracy (Test accuracy should be greater than 60%)

Step 5: Write Your Algorithm

Combine the pre-defined face and dog detector functions with the ResNet-50 based prediction model

Step 6: Test Your Algorithm

Test the algorithm with various types of user provided images and output the predicted dog breed, resembling dog breed or an error.

5. Medium post

There is an article about this project.