Crop Guard

Inspiration

Our inspiration arose from witnessing the devastating impact of bird predation on South Sudan's rural communities. The staggering 50% annual crop loss compelled us to act, recognizing the urgent need for a transformative solution to safeguard food security and livelihoods. This dire situation has left many farmers, who were previously able to feed their families, without the means to do so. Bird predation not only jeopardizes food security but also results in the loss of crucial income streams for these communities. It is our mission to provide a sustainable remedy through innovative technologies, ensuring that farmers can regain control over their livelihoods and secure a stable future for their families.

What it does

Our system serves as an advanced and intelligent bird deterrent. Utilizing a camera for bird detection, the moment a bird is identified, the system triggers ultrasonic devices. These devices emit frequencies that have the potential to distort or disorient birds through their irritating sounds, compelling them to fly away. Essentially, our technology leverages the influence of these ultrasonic deterrents to create a humane and effective solution for bird control. By seamlessly integrating camera detection, ultrasonic technology, and real-time responsiveness, our system provides a high-tech means to mitigate the impact of birds on agricultural settings, safeguarding crops and promoting sustainable farming practices.

How we built it

For the development of our solution, we implemented a proof of concept using TensorFlow.js, specifically leveraging the COCO-SSD (Common Objects in Context - Single Shot Multibox Detector) pretrained model for object detection. We crafted JavaScript functions that, when integrated into the web application, triggered an alarm whenever a bird was detected by the webcam. The user interface was designed using straightforward HTML, CSS, and JavaScript, ensuring simplicity and accessibility. The web application was then deployed on Vercel, providing a seamless and efficient platform for users. This approach allowed us to harness the power of TensorFlow.js and the capabilities of the COCO-SSD model, demonstrating the practical integration of machine learning in a real-world scenario for bird detection. The use of these technologies, along with the straightforward deployment on Vercel, underscores the effectiveness and accessibility of our innovative solution.

Challenges we ran into

We confronted numerous challenges throughout our journey. The scarcity of data in Africa posed a substantial hurdle, requiring us to navigate the complexities of acquiring and curating datasets. Learning new concepts and frameworks such as tensorflow.js added an additional layer of complexity, demanding adaptability and a steep learning curve. Setting up environments, a seemingly routine task, presented its own set of challenges, requiring meticulous attention to detail. Bugs became part of our daily problem-solving routine. Crafting an efficient function to save data proved challenging, and overcoming issues such as request failures further tested our troubleshooting skills. Despite these obstacles, our commitment to the project and collective problem-solving abilities allowed us to address each challenge methodically. In parallel, as we face the challenge of training a model to recognize bird voices amid various sounds, acquiring the necessary hardware emerged as a significant obstacle. The high cost and rarity of hardware with sufficient storage and processing power for model training added layers of complexity to our project. The search for a simple and centralized dataset containing audible and operational sounds proved to be a formidable task. The issue of hardware for our MVP was also a huge challenge.

Accomplishments that we're proud of

Our proudest achievement lies in creating a compelling demo that mirrors the envisioned system. Using a webcam as a substitute for visual sensors and laptop speakers for the ultrasound deterrence device, the demo vividly illustrates our innovative solution. Within just three days, we successfully trained the model to identify birds and trigger the deterrence mechanism, showcasing the efficiency of our vision. Equally gratifying is the seamless teamwork exhibited throughout the project lifecycle. From collaborative brainstorming sessions and thorough research to the actual coding, every team member played a vital role. This collective effort not only accelerated progress but also fostered a collaborative spirit that defines our approach. Moreover, Making it the end of the hackathon is also something we are very proud of as a team.

What we learned

Through this project, we gained valuable insights from multiple perspectives. I learned about the challenges faced by farmers dealing with bird predation – the struggle of having to stay on their farms to protect their crops, highlighting the personal sacrifices they make for their livelihoods. Traditional methods, which were once the go-to solutions, proved ineffective in addressing the issue. This realization emphasized the need for a more innovative and impactful approach. It became clear that relying solely on age-old techniques was no longer sufficient to combat the complex challenges posed by bird predation. The transformative power of leveraging modern technologies, specifically machine learning and the Internet of Things (IoT), emerged as a beacon of hope. By incorporating these advanced tools, we were able to create a smart and lasting solution. This not only showcased the potential for technology to revolutionize agriculture but also highlighted the importance of adapting to contemporary methods to address age-old problems.

What's next for Crop Guard

Next for CropGuard involves significant advancements. We plan to create a user-friendly dashboard for farmers, offering real-time insights into crop protection, bird activity, and overall field health. The dashboard will include features like pest detection summaries, weather forecasts, and actionable recommendations. Simultaneously, we are dedicated to refining our models, focusing on converting collected sounds into visual spectrograms. This innovation aims to streamline model training, making it more efficient and accurate in identifying bird threats. Looking ahead, CropGuard is poised for further innovation. We're committed to expanding the model's capabilities to adapt to various wildlife beyond birds. This includes enhancing detection features to safeguard crops from animals such as monkeys, zebras, and even elephants. Furthermore, we envision impactful partnerships, particularly with farmer cooperatives. Small-scale farmers, sharing hardware resources, will benefit from our technology, fostering community-driven solutions to bird predation. Collaborating with governmental bodies is another crucial step. By partnering with the government, we aim to integrate CropGuard into broader agricultural initiatives, contributing to national efforts for sustainable farming practices and food security.

Built With

• coco-ssd
• css
• html
• javascript
• tensorflow.js
• vercel
• webgl