Highlights and Winners of the 2023 tinyML Challenge
A heartfelt congratulations to all the winners of the 2023 tinyML Challenges! Your contributions have undoubtedly pushed the boundaries of tinyML applications in diverse fields.
Hey Community,
This week marks the grand finale of the 2023 tinyML Challenge, a journey that began in August and witnessed the participation of 91 individuals from over 20 countries. With 14 teams submitting a total of 44 innovative solutions, the challenge showcased the global enthusiasm for developing cost-effective, low-power, and reliable tinyML solutions.
This year, Seeed Studio has the opportunity to provide hardware support for participants to aid in their development efforts, including: SenseCAP K1100 x 1pcs; XIAO + Grove Base for XIAO x 1pcs + Grove Sensors x 5pcs; Micro:Bit V2.2 x 1pcs + BitMaker_V2 – JST x 1pcs + Grove Sensors x 5pcs. In total, 15 hardware kits were distributed to different applicants, with projects focusing on the three problem statements: ①Next-Gen tinyML Smart Weather Station, ②Crop disease detection, ③Wildlife monitoring.
With the judges’ evaluations and discussions, on November 27th, the online finale revealed the winners for two problem statements: Plant Disease Detection and Wildlife Monitoring. Let’s congratulate the outstanding projects and their creators!
Problem Statement: Plant Disease Detection
1. Tiny Mono – First place – Mohanasundaram SV, Purushothaman R, and Ramasamy Srinivasagan
“The agricultural sector faces several challenges in its efforts to increase production. One of the biggest challenges are diseases and insect pests that destroy crops and lead to a decline in production. This work adopts modern technological advancements in Deep learning by developing a comprehensive Tiny ML application in disease detection. The application will give farmers the knowledge and tols they need to make data-driven decisions such as disease management. Specifically, we present two classes of techniques: first, the application of image processing techniques related to ML algorithms and second, the application of deep learning in disease detection.”
2. Underdawgs – Second place – Sudharshan N, Saran Sundar S, Rohith S, and Raagav S R
“Innovative solutions, such as TinyML for early disease detection, hold promise in addressing these challenges and promoting sustainable agriculture. Our endeavor to advance plant disease detection has led us to the development of a state-of-the-art Quantized CNN. The methodology encompasses defining objectives for an efficient and cost-effective plant disease detection system, selecting hardware (XIAO ESP32S3 Sense) and acquiring a diverse dataset. A custom TinyML model is trained and optimized on Edge Impulse for hardware constraints and accuracy. The model is quantized and integrated into the hardware prototype, including components like the OV2640 camera and Lithium-Ion battery.”
Problem Statement: Wildlife Monitoring
1. AI4D Lab – First place – Fatma Issa, Jabhera Matogoro, Zephania Reuben, Ramadhani Massawe, Rogers Kalunde, Paul Mkai, Madaraka Marco Masasi, Ipyana Issah Mwaisekwa
“Wildlife conservation has always presented a myriad of challenges, and the accurate and efficient monitoring of biodiversity remains a critical concern. Traditional methods, often marked by high costs and energy consumption, are gradually being augmented and, in some cases, supplanted by the capabilities of TinyML. By leveraging TinyML models tailored to the specific needs of wildlife monitoring, ensuring that these models are lightweight, efficient, and capable of running on devices with minimal power and processing capacity. This enables real-time monitoring in remote areas while minimizing financial and environmental costs. “
2. ADN Innovators – Second place – Abhay Bhosle
“In recent years, the intersection of tinyML and wildlife monitoring has opened new frontiers in conservation and environmental research. Our device ‘Prani’ presents a comprehensive solution developed for the TinyML Challenge-03. The project utilizes the XIAO ESP32S3 microcontroller, equipped with an OV2640 camera sensor, as the core processing unit. The integration of various sensors, including the Grove Laser PM2.5 Sensor, Grove TOF Range Sensor, Grove Sunlight Sensor, and Grove Color Sensor, enhances the system’s capability to monitor both wildlife and environmental parameters concurrently.”
3. Underdawgs – Third place – Sudharshan N, Saran Sundar S, Rohith S, Ch Leela Sri, and Jayaprakash M
“Building upon the primary objective of developing efficient, lowcost, and low-power solutions for wildlife monitoring, the research aims to pioneer a paradigm shift in the intersection of technology and biodiversity conservation. The methodology encompasses defining objectives for an efficient and cost-effective wildlife monitoring system, selecting XIAO ESP32S3 Sense and acquiring a diverse dataset. A custom TinyML model is trained and optimized on Edge Impulse for hardware constraints and accuracy.”
A heartfelt congratulations to all the winners of the 2023 tinyML Challenges! Your contributions have undoubtedly pushed the boundaries of tinyML applications in diverse fields.
In conclusion, the 2023 tinyML Challenges not only showcased the incredible talent and innovation within the tinyML community but also underscored the growing significance of this field in addressing real-world challenges. The diverse range of projects, from advanced plant disease detection to effective wildlife monitoring, reflects the adaptability and versatility of tinyML solutions.
About Author
