Dog_Breed_Classifier_CNN

Dog Breed Classification Project which uses CNN's to classify real world dog images and predict their breed out of 133 breeds/classes. If supplied an image of a human face, the code will identify the resembling dog breed. Achieved an accuracy of 81% (with just one epoch!!!) using VGG-16 and Transfer learning.

I learned how to build a pipeline that can be used within a web or mobile app to process real-world, user-supplied images. Transfer learning was also used to use already trained models (VGG-16 in this case) which perform well.These models take weeks to train on high-end GPU's and hence it's not feasible to train them on PC.

Project Information

Contents

• Intro
• Step 0: Import Datasets
• Step 1: Detect Humans Accuracy-98%
• Step 2: Detect Dog Accuracy-98%
• Step 3: Create a CNN to Classify Dog Breeds (from Scratch) Accuracy-16%
• Step 4: Create a CNN to Classify Dog Breeds (using Transfer Learning) Accuracy-81%
• Step 5: Writing Own Algorithm
• Step 6: Testing Own Algorithm

Main CNN Model

I used this architecture in the step 3

and I used VGG-16 for the transfer learning in step 4. Here is the architecture of VGG-16:

Final Prediction

Future tasks to make my project stand out

1 AUGMENT THE TRAINING DATA

Augmenting the training and/or validation set might help improve model performance.

(DONE NOW)

2 TURN YOUR ALGORITHM INTO A WEB APP

Turning the code into a web app using Flask! . Planning to deploy on aws cloud.

3 OVERLAY DOG EARS ON DETECTED HUMAN HEADS

Overlay a Snapchat-like filter with dog ears on detected human heads. can determine where to place the ears through the use of the OpenCV face detector, which returns a bounding box for the face. would also like to overlay a dog nose filter, some nice tutorials for facial keypoints detection exist here .

4 ADD FUNCTIONALITY FOR DOG MUTTS

Currently, if a dog appears 51% German Shephard and 49% poodle, only the German Shephard breed is returned. The algorithm is currently guaranteed to fail for every mixed breed dog. Of course, if a dog is predicted as 99.5% Labrador, it is still worthwhile to round this to 100% and return a single breed; so, need to find a nice balance.

5 EXPERIMENT WITH MULTIPLE DOG/HUMAN DETECTORS

Perform a systematic evaluation of various methods for detecting humans and dogs in images & Provide improved methodology for the face_detector and dog_detector functions.

Libraries

The list below represents main libraries and its objects for the project.

• PyTorch (Convolutional Neural Network)
• OpenCV (Human Face Detection)
• Matplotlib (Plot Images)
• Numpy

Dataset

• Download the dog dataset
• Download the human_dataset

Accelerating the Training Process

In the training step in the Step 3 and 4, it will take too long to run so you will need to either reduce the complexity of the VGG-16 architecture or switch to running the code on a GPU or use Google Colab.

Amazon Web Services

I Used Amazon Web Services to launch a GPU instance. (This costs money!)