🤖 Multi-Project Data Science & AI Portfolio

Welcome to my portfolio repository containing 11 end-to-end Data Science, Machine Learning, and Computer Vision web applications powered by Python, OpenCV, Scikit-Learn, PyTorch, Ultralytics YOLO, and Streamlit.

---

📋 Table of Contents

Original 3 Projects (root folder):

1. Flight Price Prediction Agent
2. PAN Card Tampering & Fraud Detection App
3. Dog Breed Prediction App

Course Projects (8 subfolders): 4. Image Watermarking App 5. Traffic Sign Classification 6. Text Extraction (OCR) App 7. Plant Disease Prediction 8. Vehicle Detection & Counting 9. Face Swapping App 10. Data Science & ML Compiler App 11. Multilingual Indian Languages TTS App

---

✈️ 1. Flight Price Prediction Agent & Data Science Web App

An end-to-end Data Science and Machine Learning web application that estimates flight ticket prices in real-time, provides market analytics, and delivers intelligent booking recommendations.

• Model: Random Forest Regressor (Scikit-Learn)
• Command: streamlit run flight_app.py --server.port 2011 --server.address 127.0.0.1

---

🔍 2. PAN Card Tampering & Fraud Detection App

A Computer Vision and Image Processing web application designed to verify authenticity and detect structural tampering, modifications, or fraud in PAN card documents using SSIM and OpenCV.

• Command: streamlit run pan_app.py --server.port 2012 --server.address 127.0.0.1

---

🐶 3. Dog Breed Prediction App

A deep learning and computer vision web application that classifies dog breeds from uploaded images, provides confidence scores, temperament analysis, and care statistics.

• Model: ResNet-18 (pre-trained via timm, fine-tuned)
• Command: streamlit run dog_app.py --server.port 2013 --server.address 127.0.0.1

---

🖼️ 4. Image Watermarking App — 01_image_watermark/

Upload an image and apply a text watermark (font size, opacity, position, rotation, color) or a logo watermark (scale, position, opacity), preview side-by-side, and download the result.

• Tech: Pillow (PIL) — non-destructive RGBA compositing
• Command: streamlit run 01_image_watermark/app.py --server.port 2001

---

🚦 5. Traffic Sign Classification — 02_traffic_sign/

Upload a traffic-sign photo and get the top-3 predicted classes with confidence bars + German sign meaning. Trained on the official GTSRB dataset (12 classes, 99.5% holdout accuracy).

• Model: Custom small CNN (PyTorch), trained on CPU
• Dataset: GTSRB via HuggingFace (tanganke/gtsrb)
• Command: streamlit run 02_traffic_sign/app.py --server.port 2002

---

🔤 6. Text Extraction (OCR) App — 03_text_extraction/

Upload an image and extract all text via OCR with preprocessing toggles (grayscale / threshold / upscale), detected-region overlays, metric cards, and .txt download.

• Engine: EasyOCR (deep-learning OCR); Windows-OCR fallback
• Command: streamlit run 03_text_extraction/app.py --server.port 2003

---

🌱 7. Plant Disease Prediction — 04_plant_disease/

Upload a leaf photo and get the top-3 predicted diseases with confidence, plus treatment guidance. Trained on PlantVillage (12 disease classes, 5,940 images).

• Model: LeafNet — custom CNN (PyTorch), trained on CPU
• Dataset: PlantVillage via HuggingFace (BrandonFors/Plant-Diseases-PlantVillage-Dataset)
• Command: streamlit run 04_plant_disease/app.py --server.port 2004

---

🚗 8. Vehicle Detection & Counting — 05_vehicle_detection/

Upload an image or video; the app draws bounding boxes on vehicles (car, bus, truck, motorcycle, bicycle) and shows per-class + total counts. Videos get an annotated output download.

• Model: YOLOv8-nano (Ultralytics), pre-trained on COCO
• Command: streamlit run 05_vehicle_detection/app.py --server.port 2005

---

🔄 9. Face Swapping App — 06_face_swap/

Upload two photos (a source face + a target person); the app detects both faces, swaps the source face onto the target, and lets you download the result.

• Tech: MediaPipe face landmarks + OpenCV affine warp & blending
• Command: streamlit run 06_face_swap/app.py --server.port 2006

---

🛠️ Installation & Setup

1. Clone the Repository:

   git clone https://github.com/Vishuu77/Machine-Learning-Data-Science-Projects-.git
   cd Machine-Learning-Data-Science-Projects-

2. Install Dependencies:

   pip install -r requirements.txt

   (Each subfolder also has its own README with app-specific notes.)

3. Note on datasets: large raw datasets are excluded from this repo (see .gitignore). Run each project's train_model.py to re-download and retrain, or use the pre-trained model.pt artifacts already included.

---

📄 License

Open-source under the MIT License.