OrchardEyes

OrchardEyes

Drone-Powered Farm Health: Simple Insights, Smarter Decisions

OrchardEyes

OrchardEyes

Drone-Powered Farm Health: Simple Insights, Smarter Decisions

The problem OrchardEyes solves

Challenges in Managing Large Orchards

Monitoring Tree Health and Nutrient Levels
Effective Pest Control

Limitations of Traditional Manual Methods

-Time-Consuming and Costly
-Prone to Errors
Leads to delayed responses and inefficient resource use.

Impacts of Nutrient Mismanagement

-Poor Yields and Tree Health
-Environmental Harm and Increased Costs

Consequences of Ineffective Pest Monitoring

-Reactive Control Strategies
-Potential Crop Losses

Difficulties in Yield Prediction

-Impacts Planning and Market Strategies
-Overproduction or Underproduction
Affects profitability.

Benefits of Drone Technology

-Enhances Accuracy and Efficiency in Orchard Management
-Provides Real-Time Data for Proactive Decision-Making
-Reduces Labor and Costs
-Improves Sustainability

Challenges we ran into

Challenges Faced During the Hackathon:

1.Multiple Sampling in Yield Calculation
During our yield prediction process, we encountered an issue where the same tree was being counted multiple times because multiple frames in the video captured the same tree. This resulted in inaccurate yield calculations when running the video through our machine learning model.

Solution:
To address this, we developed a tree detection system. This system uses an additional camera to determine the tree's position, allowing the drone to align itself with the tree and capture still images rather than videos. This ensures that each tree is only captured once, providing more accurate data. Additionally, by switching to still images, we significantly reduced the required internet bandwidth compared to streaming video.

2.Limited Internet Bandwidth in Farms
Not all farms have sufficient bandwidth to transfer large amounts of data to the cloud, which could create a bottleneck and disrupt the system's performance.

Solution:
Our image-based approach, as mentioned above, helped reduce data size. We further minimized the data by focusing on key parts of the plant. Most machine learning models for detecting diseases and pests are trained on isolated images (e.g., a single leaf, flower, or fruit). Instead of capturing full-tree images, we trained our ML model to recognize various plant parts, reducing the number of images needed per tree. This allowed us to efficiently process data without overwhelming the network.

3.Power Distribution Board Track Damage
During initial drone testing, we faced an issue where the negative track of the power distribution board was damaged, disrupting the power supply to the drone's components.

Solution:
While the solution itself was simple, identifying the issue was quite challenging. After thorough troubleshooting, we identified and fixed the damaged track, ensuring stable power distribution for the drone

Tracks Applied (2)

Best Use of Auth0

Our project is designed for two main user groups: Farmers and Buyers. Farmers use the app to gain valuable insights into...Read More

Major League Hacking

Best Use of Streamlit

Our project uses streamlit for different UI components given below: chatbot UI - we have implemented Retrieval Augmented...Read More

Major League Hacking

Discussion