🧬 AMR-Watch

Predictive AMR Risk Monitoring & Response Coordination Platform

Integrated with:

---

⭐ If you like what you see, do remember to star the repository!

Yt Demon Link: Click Here

---

🧭 Introduction

AMR-Watch is a predictive public health monitoring platform that visualizes Antimicrobial Resistance (AMR) risk trends across regions and links real-time data to on-ground action

The platform combines machine learning predictions, geospatial visualization, and public health workflow automation to help municipal authorities and healthcare teams respond proactively to AMR risk hotspots.

---

Porject notebooks

# AMR-Watch Project


AMR-Watch Project/
│
├── 📜 1. chennai_wards_datamerging.ipynb
│   │
│   └── 🎯 Purpose: Data Aggregation & Geospatial Unification
│       ├── 📂 Inputs:
│       │   ├── 📍 KML files (Ward Boundaries, Flood Points, UPHCs)
│       │   ├── 📊 Excel/CSV files (Census Data, Water Quality)
│       │   └── 🗺️ Shapefiles (Drainage Networks)
│       │
│       ├── ⚙️ Processes:
│       │   ├── Merges disparate geospatial and tabular datasets
│       │   ├── Cleans and standardizes ward names and IDs
│       │   ├── Aligns all data to a common geospatial framework (ward boundaries)
│       │   └── Computes initial features like population density
│       │
│       └── 📄 Output: A preliminary GeoPackage file with unified ward-level data
│
├── 📜 2. elevation_landcover_extract.ipynb
│   │
│   └── 🎯 Purpose: Geospatial Feature Engineering
│       ├── 📂 Input: The unified GeoPackage file from the previous step
│       │
│       ├── ⚙️ Processes:
│       │   ├── Extracts mean, min, and max elevation for each ward from a Digital Elevation Model (DEM)
│       │   ├── Calculates the percentage of different land cover types (e.g., urban, water, vegetation) within each ward
│       │   └── Enriches the ward data with these new environmental features
│       │
│       └── 📄 Output: An enriched GeoPackage file (f_data.gpkg) containing comprehensive geospatial and environmental features
│
├── 📜 3. AmrRisk_prediction.ipynb
│   │
│   └── 🎯 Purpose: AMR Risk Score Calculation
│       ├── 📂 Input: The enriched GeoPackage file (f_data.gpkg)
│       │
│       ├── ⚙️ Processes:
│       │   ├── Implements a weighted formula to calculate an AMR risk score for each ward
│       │   ├── Considers factors like microbial load, flood risk, heavy metal concentration, and health infrastructure
│       │   └── Assigns a final, quantifiable risk score to each ward
│       │
│       └── 📄 Output: The final GeoPackage (chennai_ward_amr_risk.gpkg) with an added AMR_risk_score column, ready for the backend server
│
└── 🤖 Modeling Branch (Utilizes data from the pipeline)
    │
    ├── 📜 4. flood_prediction_model.ipynb
    │   │
    │   └── 🎯 Purpose: Train Flood Prediction Model
    │       ├── 📂 Inputs: Historical data on rainfall, temperature, humidity, river discharge, etc.
    │       │
    │       ├── ⚙️ Processes:
    │       │   ├── Preprocesses and cleans the historical flood dataset
    │       │   ├── Trains a machine learning model (e.g., XGBoost, RandomForest) to predict flood probability
    │       │   └── Serializes the trained model into a pickle file
    │       │
    │       └── 📄 Output: A trained model file (flood_batata.pkl)
    │
    └── 📜 5. finetuned_vgg_diseasepred.ipynb
        │
        └── 🎯 Purpose: Train Image-based Disease Classifier
            ├── 📂 Inputs: A labeled dataset of skin disease images (e.g., ringworm, eczema)
            │
            ├── ⚙️ Processes:
            │   ├── Implements data augmentation to improve model robustness
            │   ├── Utilizes transfer learning by fine-tuning a pre-trained VGG model on the specific disease dataset
            │   └── Saves the trained model weights
            │
            └── 📄 Output: A trained deep learning model file for disease classification

🌍 Core Modules

1️⃣ Map Layer Toggles for AMR Risk Scores 🗺️

Interactive overlays let officials correlate AMR risk with key environmental and infrastructural factors.

Buttons: Flood_data

⚙️ Actions & Integration

Clicking a layer overlays infrastructure data over AMR hotspots.

Random Forest–based AMR risk score calculation.

🧠 Interpretation

The resulting AMR Risk Score provides a continuous numeric measure of relative antimicrobial resistance risk across Chennai’s wards. Higher values indicate greater risk due to microbial contamination, flood exposure, or environmental stressors; lower values represent better resilience and healthcare access.

🧾 Automated Task Creation

Clicking a high-risk zone triggers an automated workflow:

Task Title: New Infrastructure Task: [Ward Name]

Task Type: Auto-selected — AMR Risk Scores, Live Flood Prediction

Pre-filled Data: AMR risk score, coordinates, and causal factors from open sources.

Assignment: Dropdown list of municipal response teams.

Priority Selector: High / Medium / Low

✅ This pipeline enables real-time decision support for flood-prone and infection-vulnerable communities.

---

3️⃣ Disease Classifier 🔬

20251012-091159_TId2bAxK.mp4

Transforms field-level image data into verified, actionable reports.

Button: Upload Image

Actions & Integration:

• Field officers upload microscopy images from mobile devices.
• Model predicts disease, e.g.,

  “Predicted Disease: Ringworm, Confidence: 87%”

• Clicking Log Suspected Case & Notify Officer
  • Auto-filled Data: Pathogen, confidence, timestamp, GPS.
  • User Inputs: Source, patient ID (anonymized), field notes.
  • Follow-up Flag: Checkbox for urgent epidemiological investigation.
  • Submit: Sends immediate alerts to regional health officers and logs the case to the epidemiological map.

Method:

• Get pre trainer model on skin-diseases from hugging face.

• Prepare custom dataset -> finetune

• Model VGG16 + AutoEncoders

  • Accuracy: 93.4%.

  

  • Inputs: classes of own custom data.

  

  • Outputs: Disease-predicted with ai companion.

  

  4️⃣ Flood Prediction Model 🌐☁️

  🧾 Dataset Summary

Feature	Missing Values
Latitude	0
Longitude	0
Rainfall (mm)	0
Temperature (°C)	0
Humidity (%)	0
River Discharge (m³/s)	0
Water Level (m)	0
Elevation (m)	0
Land Cover	0
Soil Type	0
Population Density	0
Infrastructure	0
Historical Floods	0
Flood Occurred	0

✅ No missing values detected across all features — dataset is clean and ready for model training.

Model Summary

• RandomForest Regressor.
• Accuracy - 95%.
• Clicking on a region opens to plan preventive actions, such as deploying resources to high-risk zones or issuing public health advisories.

Live Data Integration

This module integrates real-time global weather data to provide predictive insights on environmental conditions that may influence AMR risk.

Features & Integration:

• Displays current global weather patterns using temperature, rainfall, and humidity overlays.

• Highlights regions where extreme weather may contribute to AMR hotspots.

• Openweather live data for all wards

  Elevation and Landcover Integration

  

--

🧠 Tech Stack

Category	Technologies
Frontend	React+vite, Plotly, Folium
Backend	Python, TensorFlow, Keras
Data Handling	Pandas, NumPy
Visualization	Matplotlib, Seaborn
Integration	API, REST Endpoints
Deployment	Docker

---

👩‍💻 Our Team

🏆 IBM Z Datathon 2025 — Project by Shooting Star Foundation

       

Team Name: Shooting Star Foundation 🌠
Event: IBM Z Datathon 2025
Theme: Predictive AMR Surveillance and Public Health Response
Institution: [Your College / Organization Name Here]

---

💡 Team Members

Name	Role	Focus Area
Ipsita Kar	Skin Disease Integration and Github
Ganesh	Data Engineer	AMR Risk Model & Data Pipeline
Sanjeev	UI/UX Developer	Global Temperature & Integration
Sanjeev	PPT	API for Weather Prediction and PPT

---

🌟 Proudly developed by the Shooting Star Foundation during the IBM Z Datathon 2025!

---

📊 Features

✅ Real-time AMR risk prediction
✅ Layer-wise environmental correlation
✅ Disease classification from field samples
✅ End-to-end response tracking dashboard
✅Global Temperature and weather Alert System ( For Flood Prone areas showing Possibilities )

---

🚀 Future Enhancements

• Integration with satellite rainfall prediction APIs
• Improved explainable AI visualization for policy-level insights
• Offline-first mobile interface for field data collection
• API-based cross-regional health data exchange

---

⭐ Do remember to star the repository if you like what you see!

---

🩺 Built with ❤️ to strengthen data-driven public health infrastructure.