AgriSmart
Smart Kheti, Zyada Fasal – AI ke Saang!
What’s your problem statement?
Social Good / Llama Impact Hack:
Science and Innovation - Humanity has been struggling with the big problems for centuries - climate, hunger, health, agriculture, accessibility etc. Can the rise of AI usher in solutions for the greater good?
The problem AgriSmart solves
Our product empowers farmers by providing AI-driven tools that address key agricultural challenges. Here’s how it makes farming tasks easier, safer, and more sustainable:
Knowledge Access: Farmers can now access agricultural insights through an LLM Chatbot trained on horticulture and agriculture textbooks, available in multiple languages. This knowledge boosts informed decision-making, particularly for resource-limited farmers.
Fertilizer and Crop Recommendations: With customized Fertilizer and Crop Recommenders, farmers receive optimal suggestions based on soil and climate conditions, reducing trial and error and conserving resources.
Predictive Insights: The Crop Yield Predictor and Weather Forecasting tools enable farmers to anticipate crop yields and adjust to changing weather conditions, minimizing crop loss and enhancing planning.
Crop Health & Safety: The Disease Detector ensures early identification of plant diseases, protecting crops from widespread damage. Additionally, Livestock and Intruder Detection enhance farm safety by monitoring livestock and securing farm boundaries.
These tools help farmers save time, cut costs, and achieve higher, safer yields, promoting resilience and sustainability in agriculture.
Challenges we ran into
Render Memory Limit Exceeded: We encountered memory limitations on Render’s hosting platform, impacting deployment. To manage memory effectively, we optimized model and API configurations, ensuring smooth performance without exceeding limits.
YOLO Model Compatibility: While attempting to implement YOLOv11, we faced repeated crashes during the build phase. Switching to YOLOv8 provided a more stable solution, allowing us to achieve reliable object detection without sacrificing accuracy.
Intruder Detection Deployment: Deploying the video-based intruder detection on Hugging Face with JavaScript was complex and posed integration issues. We resolved this by refining the deployment process and leveraging async processing to handle video data smoothly.
These solutions allowed us to overcome technical hurdles and ensure robust performance across all features.
Tracks Applied (1)