Team Reference Number- 5E
Team Number- 54
Team Members- Pahwa Ronak, Chidambaram Aditya Somasundaram, Sunkara Bhargavi, Banerjee Mohor
Project Name- RottenAI
Inspiration
The inspiration behind RottenAI stems from a critical and growing concern: food waste. In a world where millions face food scarcity, the excessive amount of food waste in households due to spoilage is not only an environmental concern but also a moral one. RottenAI seeks to address this issue by leveraging technology to help individuals make informed decisions about their food. By assessing the condition of food items that are presumed to be waste, RottenAI aims to minimise unnecessary disposal, encouraging more sustainable and conscious food consumption and management practices.
What it does
RottenAI is an AI chatbot that gives you useful advice on how to utilise "supposed" food waste effectively.
All the user has to do is upload an image of the food that has gone bad to the bot and key in some additional details about the food (like when they bought it) into the prompt bar.
In addition to the information provided by the user through prompts- image and user prompt- the bot also uses 4 sensors (hardware connected to the pc and mounted on an arduino)- temperature, humidity, water and moisture to give the most optimised advice on whether the food is edible or not and what is the best action (cooking, composting, etc) that can be done with it.
But wait, it is not just a text-generation bot that gives some advice.
It is also a "chatbot" to which the user can shoot follow-up questions and have an entire conversation with about the food item they uploaded.
The user can ask it for recipes that can still be made using the food item, how to make compost from it, more about its health dangers, etc.
Thus, this bot is indeed an impactful first step towards reducing food waste and encouraging better food waste management practices among common consumers like you and me!
How we built it
To construct the conversational chatbot interface of RottenAI, we leveraged the gemini-pro-vision model, a foundational model known for its efficacy in handling multimodal tasks, including processing both text and image inputs to generate textual outputs. Our development stack included Flask and Python for the backend, alongside various JavaScript frameworks to create a dynamic and responsive frontend.
A key aspect of RottenAI's functionality is its ability to integrate real-time data from physical sensors. We achieved this by connecting four environmental sensors—measuring temperature, humidity, water, and moisture—to an Arduino board. These sensors are crucial for assessing the condition of food items in a nuanced way. We developed a system to take serial inputs from these sensors connected to the PC, processing the data in real time to inform the AI's advice. This integration required meticulous calibration and testing to ensure accuracy and reliability in the sensor data collection process.
Our solution's architecture was designed to be modular, allowing for seamless communication between the hardware inputs and the software logic. By utilizing Python's serial library to read data from the Arduino, we were able to feed environmental conditions directly into our AI model. This enabled RottenAI to provide contextually relevant advice based on visual analysis of the uploaded food images, quantitative data from the sensors and additional details provided by the user as a natural language prompt.
Challenges we ran into
In developing our project, we faced three major technical challenges. First, integrating hardware sensors with software required precise calibration and extensive testing to ensure seamless data collection. Second, utilising the Gemini API, a new entrant in the generative AI field, presented a steep learning curve due to its sparse documentation, pushing us to innovate and troubleshoot independently. Lastly, managing data flow between Python and JavaScript demanded mastery of inter-language communication, highlighting the need for adaptability in handling syntax and execution differences. These challenges not only tested our technical skills but also our resilience, driving us to find creative solutions and advance our project.
Accomplishments that we're proud of
Throughout the development of RottenAI, we learned the importance of interdisciplinary collaboration, combining expertise in AI, IoT, and environmental science to create a holistic solution. We also gained insights into the challenges of processing and interpreting real-world data, such as the variability in food spoilage indicators and user interaction patterns
What we learned
In developing RottenAI, we gained insights into the integration of AI and IoT for environmental solutions, emphasising the importance of interdisciplinary approaches and technical flexibility. Overcoming challenges in hardware-software integration and AI utilisation refined our problem-solving capabilities, demonstrating the potential of technology in addressing sustainability issues.
What's next for RottenAI
Looking ahead, we aim to enhance RottenAI's capabilities by expanding its database for greater accuracy in spoilage assessment and including more diverse food items. We plan to integrate additional sensors (like gas concentration sensor, pH sensor, etc) for better environmental monitoring and explore partnerships with food-related organisations to reach a broader audience. Continuous improvement of the chatbot's conversational AI will ensure more engaging and informative interactions. Ultimately, our goal is to make RottenAI a cornerstone in household food management, significantly reducing food waste on a global scale and promoting sustainability.
Reference for video - https://www.youtube.com/watch?v=TVP3j7_W7og logo - https://www.freepik.com/premium-vector/illustration-joyful-ripe-tomato-sad-rotten-tomato-expressive-character-with-arms-legs_37982883.htm
Log in or sign up for Devpost to join the conversation.