-
Objective
- Build an Convolutional Neural Network capable of classifying images from a dataset containg driving licenses, social security cards amd miscellaneous items.
- This could be useful for companies requiring KYC documentation.
-
Technology
- Python, Scikit-learn, TensorFlow, Keras, Pandas, Numpy, Flask
-
Metrcs
- Accuracy
-
Deployment
- Deploy model in a scalable way so that business decisions can be taken in near real time in assessing riskiness of a transaction
-
Approach
- Data Loading
- Data Preprocessing
- Build a baseline CNN model to classify images
- Data Augmentation
- Predictions
- Model Deployment
python Engine.py
Train - 0
Predict - 1
Deploy - 2
Enter your value: 0
Number of images for training: 600
Preprocessing has begun...
Found 600 files belonging to 3 classes.
Using 480 files for training.
....
Found 600 files belonging to 3 classes.
Using 120 files for validation.
Class Names: ['driving_license', 'others', 'social_security']
Data loading has completed...
Preprocessing is complete...
Enter your value: 1
Number of images for testing 150
Found 150 files belonging to 3 classes.
Enter your value: 2
1/1 [==============================] - 0s 333ms/step
127.0.0.1 - - [11/Dec/2022 14:02:30] "POST /get-image-class HTTP/1.1" 200 -
Real Time Prediction
