Infinage / Age-Prediction

End-to-end machine learning project showcasing age prediction with a comprehensive pipeline, including TFLite conversion, live UI with TensorFlow.js, and Dockerized development environment.

Age Detection

The goal of this project is to explore the full end to end life cycle of a Machine learning project. Although the model predictions aren't great, the pipeline to build, train and deploy has been tested to work well.

A live inference demo is deployed on

1. Netlify.
2. AWS with custom Subdomain.

TODO

1. Train a simple age detection model
2. Convert trained model to TFLite
3. Using Opencv JS & TFJS create a live age detection UI
4. Dockerize development environment
5. Deploy the front end only code
6. Mobile friendly inference
7. Refactor code base to use async await instead of event listeners?
8. Improve age prediction model accuracy
9. Explore options to train and use Yolo for face detection

Useful links

• Age prediction dataset: https://www.kaggle.com/datasets/mariafrenti/age-prediction
• Face detection with Open CV JS: https://docs.opencv.org/3.4/d2/d99/tutorial_js_face_detection.html
• Yolo training with Keras_cv: https://keras.io/examples/vision/yolov8/
• Face detection Dataset: https://www.kaggle.com/datasets/fareselmenshawii/face-detection-dataset/data
• Convert Model to TFLite: https://www.tensorflow.org/lite/models/convert
• Using TFLite with TFJS: https://js.tensorflow.org/api_tflite/0.0.1-alpha.10/
• TFJS Boiler plate copied from Google: https://codelabs.developers.google.com/codelabs/tensorflowjs-object-detection#7
• TFlite Sample Inference: https://codepen.io/jinjingforever/pen/xxgWRaE?editors=1010
• Open CV JS Object detection: https://docs.opencv.org/3.4/df/d6c/tutorial_js_face_detection_camera.html