Inspiration Natural disasters require fast decisions and coordinated responses. Damage Assessment And Dispatchment Unit was created to help assess disaster images quickly and recommend appropriate emergency response teams.
What it does The application allows users to upload an image and classify it as one of four disaster types:
Cyclone Earthquake Flood Wildfire After classification, the system recommends response teams such as firefighters, medics, rescue teams, water rescue, and police. Users can either review each dispatch request with Allow or Decline, or use Assess & Dispatch to call the recommended teams immediately.
How we built it Built the user interface with Streamlit. Used PyTorch and TorchVision for image classification. Trained a disaster classification model using labeled disaster images. Added image preprocessing, resizing, normalization, and model inference. Created separate response functions for emergency teams. Added session state to manage results, dispatch approvals, and clearing the current assessment. Archived processed images for later review. Challenges we ran into Connecting the trained model to the Streamlit interface. Handling uploaded images in different formats. Managing Streamlit reruns and preserving dispatch decisions. Preventing emergency functions from running before the user approves them. Supporting both manual approval and immediate dispatch modes. Keeping function references stable across Streamlit reruns. Accomplishments that we're proud of Built a working disaster image classification application. Added a clear approval workflow before dispatching response teams. Added an immediate Assess & Dispatch option. Connected different disaster types to appropriate response teams. Added a readable dark interface with clear status messages. Added image archiving and a Clear button for repeated assessments. What we learned We learned how to integrate a PyTorch model into a Streamlit application and how to design a workflow around model predictions. We also learned the importance of session state in Streamlit, especially when handling buttons, reruns, uploaded files, and multi-step user decisions.
What's next for DamageAssesmentAndDispatchmentUnit Improve model accuracy with a larger and more diverse dataset. Add confidence scores to predictions. Support multiple images in one assessment. Add location and severity detection. Connect dispatch functions to real emergency service APIs. Add user authentication and secure access. Store assessments in a database. Add an incident history dashboard. Improve accessibility and mobile responsiveness.
Log in or sign up for Devpost to join the conversation.