Banana AI
Highlights
- Object detection for advanced monitoring
- Object classification with deep learning
- AI-based software module for image extraction
- Comprehensive reports on the plant’s condition
Challenge
The client has reached out to SoftTeco with the aim of developing an innovative AI-driven solution for monitoring, controlling, and observing the cultivation of banana trees in greenhouses and plantations to automate the routine tasks of agronomists and promptly detect issues.
The client’s idea revolved around developing a computer vision system capable of automatically inspecting banana seedling leaves for signs of damage and characterizing them. However, the main challenge was the lack of suitable data to train the algorithm.
Solution
SoftTeco employed object detectors and deep learning techniques to create and train a model capable of identifying damaged banana leaves. Our specialists have scrupulously collected the essential dataset required for model training.
Next, Softteco designed a specialized software module aimed at inspecting and categorizing the possible leaf damage. Following the analysis of the captured images, this module generates a comprehensive report that offers a detailed assessment of the plant’s condition.
Tech Stack
Components
Python
Pytorch
OpenCV
Pandas
NumPy
Streamlit
How it works
At our recommendation, the client strategically placed object-detecting cameras within the greenhouse, primarily positioning them above and to the side, to capture periodic snapshots of banana seedlings. The client can adjust the frequency of these image captures according to their preferences.
Using Python and PyTorch, SoftTeco’s team designed and implemented a software module that extracts images of detected leaves and forwards them to the classifier model for further in-depth analysis. SoftTeco’s specialists developed and fine-tuned the classifier model, ensuring it meets the unique requirements of this project.
Once the classifier model receives the images, it engages in a systematic classification process, meticulously distinguishing between damaged and healthy leaves. After this analysis, the model generates an exhaustive PDF report that provides a comprehensive overview of the plant’s condition. Furthermore, it offers valuable care recommendations to guide agronomists in optimizing the growth of banana seedlings.
Results
SoftTeco has not only built and fine-tuned the classifying model but has done so with meticulous attention to detail, aligning closely with the project’s unique requirements. The client is fully satisfied with the result and continues to work with us on further projects.
Have a project in mind?
Let us know what kind of software solution you need, and our specialists will provide an estimate cost and deadline.
