The problem KAVACH solves

Human-wildlife conflict poses a significant safety challenge for communities located near forest and wildlife habitats. Existing monitoring and response mechanisms can be difficult to manage across large areas, often resulting in delayed awareness and response when wildlife enters human-populated zones.

KAVACH addresses this challenge by providing a centralized platform for monitoring wildlife activity, identifying potential conflict-prone areas, and supporting timely decision-making.

The system can be used to:

Monitor wildlife activity across vulnerable areas.
Identify locations with a higher likelihood of human-wildlife encounters.
Provide timely alerts to relevant authorities and stakeholders.
Visualize wildlife incidents and risk-prone zones on a geographical interface.
Maintain historical information to identify recurring conflict patterns.
Support forest officials in planning preventive measures and allocating resources more effectively.

By bringing monitoring, location-based information, alerts, and historical insights into a single platform, KAVACH aims to make wildlife conflict management more proactive, organized, and effective.

Challenges we ran into

One of the key challenges was training YOLO for wildlife detection. Finding suitable datasets for Indian wildlife species was difficult, as many publicly available datasets did not sufficiently match the species relevant to our use case. Preparing suitable training data and getting reliable detection performance required significant experimentation.

We also faced challenges in sourcing location-based datasets that could support meaningful risk intelligence for wildlife conflict-prone areas.

Another major hurdle was integrating the complete technology stack. Connecting the trained YOLO model with our ML-based risk intelligence, backend, and frontend required careful handling of data flow and APIs so that detection results could be processed and displayed correctly.

We overcame these challenges through iterative dataset preparation, model training and testing, and step-by-step integration of the YOLO, ML, backend, and frontend components into a single working system