Building a Flask app on Image Classification of Dog/Cat Dataset implemented by Convolutional Neural Network (CNN)
Ayoub Berdeddouch
This is the project that we finished after ML ZoomCamp by @ Alexey Grigorev
INTRODUCTION
1. The Dog vs. Cat Dataset
Dogs vs. Cats dataset provided by Microsoft Research contains 25,000 images of dogs and cats with the labels
- 1 = dog
- 0 = cat
2. Project goals
-
Building a deep neural network using TensorFlow & Transfer Learning to classify dogs and cats images.
-
Making a Flask application so user can upload their photos and receive the prediction.
3. Project plan
During this project, we tackled these steps:
A. Build the model
B. Build the Flask app
Front end
- HTML
- How to connect frontend to backend
- How to draw a number on HTML
- How to make UI looks good
Back end
- Flask
- How to set up Flask
- How to handle backend error
- How to make real-time prediction
- Combine the model with the app
HOW IT WORK: CONVOLUTIONAL NEURAL NETWORK (CNN)
In deep learning, a convolutional neural network (CNN, or ConvNet) is a class of deep neural networks, most commonly applied to analyzing visual imagery. (Wiki)
For this project, we used pre-trained model MobileNetV2 from keras. MobileNetV2 is a model that was trained on a large dataset to solve a similar problem to this project, so it will help us to save lots of time on buiding low-level layers and focus on the application.
Note: You can learn more about CNN architecture here chapter 8.
MODEL PERFOMANCE SUMARY
Our model has the accuracy of 99 % for the train dataset and 98.67 % for the test dataset.
FLASK APPLICATION
Homepage
Example of results
