Amidst the ongoing war against diseases triggered by climate change, one researcher’s work lies in the conjunction of AI technologies and agriculture. Chidiebere Nwaneto, a computer science postgraduate alumnus from the University of Lagos, developed algorithms for farmers in West Africa suffering due to Taro Leaf Blight (TLB), which triggers food insecurity and affects several parts of the world. His unique efforts, which blend Deep Learning Algorithms (DL) and smartphone technologies, enable the detection of these diseases within agricultural fields for timely intervention.
Chidiebere Nwaneto’s Key Contributions
1. Revolutionizing Early Disease Detection
Nwaneto’s 2024 publication in Smart Agricultural Technology showcased his deep learning models to detect TLB and achieve 74% accuracy using a Vision Transformer (ViT) model; a feat unprecedented by other CNNs like VGG16 (56%) and ResNet50 (36%). His model’s subtle ‘water-soaked spots’ detection accomplishes early TLB detection, halting its proliferation and averting potential 95% yield losses. Nwaneto received notable recognition through award nominations, including featuring in Elsevier’s top sustainability journals and congruently gaining pilot adoption across several farms in Nigeria and Ghana.
2. The TLB-Detector Android Application: AI at the Fingertips of Farmers
Innovation: Nwaneto’s team, Dr. Chika Yinka-Banjo, and Ogban Ugot improved the already-existing object detection model by fine-tuning YOLOv8 on a dataset of 13,887 taro leaf images. The model achieved 85.7% accuracy in mAP, which was 15-20% more than competitive models.
Real-World Tool: The Android app is freely accessible and can be used in low-RAM smartphones. It can provide real-time diagnoses. Farmers can scan the leaves and get instant reports on what stage of infection it is at (early/mid/late blight).
Impact Metrics:
- 87% of the field-test farmers from 42 farmers showed user-friendliness of the app.
- 75% of the mid-stage infection could be accurately detected which made timely biocontrol possible.
- Dataset & Code: Publicly available on Mendeley and GitHub.
3. Developing the First Taro Leaf Blight (TLB) Open Image Dataset in Africa
Challenge: To train AI models, you need data tailored to a specific location. At the time of our research, we could not find any TLB Disease dataset in any repository. The infamed PlantVillage dataset does not contain any TLB variants from Western Africa.
Solution: Nwaneto’s extended team, which includes Prof. Thompson Annor (Ghana) and Dr. Oby Umeugochukwu (Germany), gathered and annotated 13,887 images based on different infection levels while also considering the climate and soil variability of the West African region.
Global Recognition: This dataset is fast becoming a standard for crop AI and has been cited in reputable international and local journals.
Reason Behind This Work That Impact Lives
Food Security: Taro is a diet for more than 300 million people in Western Africa, and Nwaneto’s tools can prevent more than 30% and up to 50% yield loss. Boosting farmers’ confidence when planning to embark on Taro cultivation.
Economic Equity: The smallholder farmers stand to save big costs on fungicides because of early biocontrol.
These models control the carbon footprint and cloud usage, qualifying them as sustainable AI.
Want To Explore The Tech?
View the project’s landing page here
Read the Research papers:
| Publisher | Journal | Title | Authors |
| Elsevier | Franklin Open | An object detection solution for early detection of taro leaf blight disease in the West African sub-region | Chidiebere B Nwaneto, Chika Yinka-Banjo, Ogban Ugot |
| Elsevier | Smart Agricultural Technology | Early Detection of the Taro Leaf Blight Disease in the West African Sub-Region Using Deep Image Classification Models | Chidiebere Nwaneto, Chika Yinka-Banjo, Ogban-Asuquo Ugot, Thompson Annor, Obiageli Umeugochukwu |
| Ife Journal of Science | Ife Journal of Science | Harnessing deep learning algorithms for early plant disease detection: A comparative study and evaluation between SSD (Mobilenet_v2 and Mobilenet_v3) and CNN model | Chidiebere B. Nwaneto, C Yinka-Banjo |
| Elsevier | Data In Brief (DIB) | Image dataset of Taro Leaf Blight disease collected from the West African Sub-Region | Chidiebere Nwaneto, Chika Yinka-Banjo, Ogban-Asuquo Ugot, Obiageli Umeugochukwu, Thompson Annor |
About Chidiebere Nwaneto
Chidiebere Nwaneto is a Software Engineer/Architect with over 13 years of experience building enterprise systems at scale. He obtained his Master’s Degree in computer science from the University of Lagos, where his research in computer science focuses on applying AI towards achieving sustainability in agriculture. His studies delve into cost-efficient, high-benefit approaches for smallholder farmers as part of the IDRC-funded RAINCA consortium. With several publications in Q1 journals, he is known as an active advocate for open-source AI. Reach out to him via his LinkedIn account. This initiative is under the Responsible Artificial Intelligence Network for Climate Action in Africa (RAINCA), which is funded by the IDRC (Grant 109705-001/002).
