Portfolio Machine Learning Models

This repository contains machine learning models trained and exported for live interactive demos on my portfolio website. Each model is showcased with a web-based demo card, allowing users to try them out in real time.

---

🚀 Models Overview

1. Sentiment Analysis (TinyBERT)

• Description: Classifies user-input text as positive or negative sentiment.
• Tech: Fine-tuned TinyBERT model.
• Demo: Users can type any sentence and instantly see the sentiment prediction.

2. Doodle Classification

• Description: Recognizes hand-drawn doodles using a model trained on the Google QuickDraw dataset.
• Tech: Convolutional Neural Network (CNN).
• Demo: Users can draw on a canvas, and the model predicts the doodle category.

---

📁 Project Structure

sentiment/         # Sentiment analysis model (TinyBERT)
doodle/            # Doodle classification model (QuickDraw CNN)
utils/             # Utility scripts (preprocessing, conversion, etc.)
outputs/           # Exported models and checkpoints

---

🛠️ Setup & Training

1. Clone the repository:

   git clone <repo-url>
   cd portfolio_models

2. Install dependencies:

   uv sync

3. Train models:

   • Sentiment: See sentiment/train_lstm_sentiment.py
   • Doodle: See doodle/train_cnn.py

4. Export models for web:

   • Use script in utils/convert_to_tfjs.py as needed.

---

🧩 Scripts Overview

Doodle Scripts (doodle/)

• quickdraw_to_png.py: Converts QuickDraw .ndjson files to PNG images for training.
• split_png_dataset.py: Splits the PNG dataset into training, validation, and test sets.
• train_cnn.py: Trains the CNN model for doodle classification.

Sentiment Scripts (sentiment/)

• prep_imdb_data.py: Prepares and splits the IMDB dataset for sentiment analysis.
• train_lstm_sentiment.py: Trains an LSTM-based sentiment analysis model.

Utility Scripts (utils/)

• convert_to_tfjs.py: Converts trained models to TensorFlow.js format for web deployment.
• preprocess_doodle.py: Preprocessing utilities for doodle data.

---

🌐 Web Integration

• Models are exported in web-friendly formats (TensorFlow.js).
• Each model is integrated into a card on the portfolio website for live demo and interaction.

---

📊 Performance

• Sentiment Analysis: [80%] for 2 divisional classes
• Doodle Classification: [63%] for 50 divisional classes

---

📄 License

MIT License

---

🙋‍♂️ Author

Anoop