How Labellerr Empowered Oishii's AI Advancements in Vertical Farming

AI in agriculture is solving problems like labor shortages, crop quality, and inefficiencies in harvesting.

In vertical farming, it is even more important because it helps farmers make the most of limited space and controlled environments. Automation is becoming a key part of vertical farming, improving productivity and reducing costs.

AI and robotics are transforming agriculture by solving problems like labor shortages, crop quality, and harvesting inefficiencies. Farmers are increasingly turning to automation to improve productivity and reduce costs.

Studies show that the global Agricultural Robots Market is experiencing prominent growth, with an estimated value projected to reach USD 51.0 billion by 2029 from the 2024 valuation of USD 16.6 billion, indicating a significant Compound Annual Growth Rate (CAGR) of 25.2%. 

In vertical farming, robotics plays a vital role in automating tasks like fruit harvesting. In this project, image segmentation is essential to train autonomous robots that work in vertical farms.

These robots need to detect the ripeness of strawberries and find the exact points on the stem to cut them accurately. This precision ensures ripe berries are harvested without damage, and unripe fruit is left untouched.

Image segmentation allows the AI model to label and understand these details, making it the foundation for the robot’s performance.

Accurate image segmentation is vital for creating reliable AI models in vertical farming. When the data is properly labeled, the model can quickly identify ripe berries, locate stems, and make precise cuts.

Poor segmentation could result in errors, such as cutting unripe berries or damaging the plant. By focusing on segmentation, this project aims to improve the efficiency of strawberry harvesting, reduce waste, and boost productivity.

Autonomous robots with these capabilities will help farmers save time, reduce manual labor, and increase the quality of their harvest.

About the Customer

Oishii is an organization in the vertical farming sector, focused on using robots to make farming smarter and more efficient.

They are committed to using AI and robotics to solve challenges in farming, such as maximizing productivity, reducing manual labor, and ensuring precise harvesting of crops.

Their goal is to create innovative solutions that help farmers adopt modern techniques and stay competitive in the evolving agricultural landscape.

For this specific project, the customer aims to develop an AI model that enables autonomous robots to identify and harvest strawberries with precision.

In vertical farming, where plants are stacked in layers and space is tight, robots need to detect ripe strawberries and find the exact cutting points on the stems.

This level of accuracy will help the robots harvest strawberries efficiently without damaging the fruit or plant.

By achieving this goal, the customer seeks to improve the overall harvesting process, save time, and reduce labor costs, making strawberry farming more sustainable and scalable.

Data Volume and Accuracy Challenge

Data Complexity

Labeling strawberry images in vertical farming presents unique challenges. The ripeness levels of strawberries vary, so annotators must carefully label ripe, unripe, and overripe berries.

In vertical farms, where plants are stacked close together, strawberries can overlap or be partially hidden by leaves. This makes it harder to identify and label them. Artificial lighting, often used in vertical farming, also creates shadows and reflections that can affect how objects like berries, stems, and calyx appear in images.

The customer required precise annotations for the berries, their calyx (the leafy top), and stems. This level of detail was essential to train the AI model to detect ripeness and cutting points.

Volume of Data

The project involves handling a large dataset with thousands of images. Each image contains multiple strawberries that need detailed labeling, resulting in tens of thousands of data points to be annotated.

Managing such a high volume of data is a significant challenge, especially when aiming to maintain consistency and quality across the annotations.

Accuracy Requirements

High accuracy was critical for the success of the project. The annotated data guided the robot in identifying ripe strawberries and determining precise stem-cutting points.

Any labeling errors could lead to problems, like cutting unripe berries or damaging plants, which would lower efficiency and increase waste.

Precise and consistent annotations were needed to ensure the AI model worked effectively in the vertical farming environment.

How Labellerr Stepped In

Tailored Solution

Labellerrcustomers customized its annotation platform to meet the specific needs of the customer. The platform was configured to handle the complexities of strawberry segmentation, ensuring accurate and efficient labeling.

Labellerr incorporated various annotation types to match the project’s requirements:

• Bounding Boxes: Used to detect the ripeness of strawberries by marking their overall shape and location in the image.
• Polygons (Masks): Applied to outline the precise edges of berries, calyx, and stems. This detailed segmentation allowed the AI model to differentiate between these components accurately.
• Key Points: Placed on specific parts of the stems to indicate the exact cutting points, enabling the robot to perform precise and clean cuts.

Advanced Features

Labellerr’s platform included advanced features that made the annotation process faster and more accurate:

• AI-Assisted Annotation: The platform used AI to pre-label data, suggesting bounding boxes, polygons, and key points. This reduced the manual effort required by the annotators and sped up the overall process.
• Quality Assurance Mechanisms: Features like confidence scoring helped identify areas where labels needed review, ensuring high accuracy. Anomaly detection flagged inconsistencies in the annotations, maintaining uniformity across the dataset.

Workflow Integration

Labellerr’s platform seamlessly integrated with the customer’s existing data pipelines.

The platform supported smooth uploading of images, annotation processes, and exporting of labeled data in formats compatible with the customer’s machine learning workflows.

This integration allowed the customer to process data without interruptions and easily incorporate it into their AI development pipeline.

By tailoring the platform, leveraging advanced features, and ensuring seamless integration, Labellerr enabled the customer to overcome the challenges of data complexity, volume, and accuracy, paving the way for the successful development of their AI model.

Results and Impact

Improved Annotation Efficiency

Labellerr significantly improved the efficiency of the annotation process by automating repetitive tasks and streamlining workflows.

With AI-assisted tools, the customer was able to complete annotations in 70% less time compared to manual methods.

Tasks that previously took weeks were reduced to just a few days, allowing the customer to focus on training and refining their AI model. This efficiency not only saved time but also reduced operational costs.

Enhanced Model Performance

Precise annotations provided by Labellerr helped the AI model perform at a much higher level.

The robot could accurately detect the ripeness of strawberries and identify the exact points on the stem for cutting. As a result, the robotic harvesting process became more reliable and precise, reducing errors such as:

• Avoiding Damage: The robot successfully avoided cutting unripe berries or damaging the plant.
• Improved Precision: The AI model’s enhanced understanding of ripeness levels ensured that only the ripest berries were harvested.

This improvement in model performance directly contributed to better harvest quality and reduced waste.

Scalability

Labellerr’s tools enabled the customer to handle large datasets effortlessly, ensuring the project could scale for future needs.

The platform managed thousands of images efficiently and prepared the dataset for additional crops or expanded use cases.

With Labellerr, the customer can now scale their annotation process to accommodate larger datasets, paving the way for more advanced applications in smart farming and robotics.

By saving time, improving model accuracy, and supporting scalability, Labellerr delivered a solution that empowered the customers to achieve their goals efficiently and effectively.

Conclusion

Labellerr’s automated annotation and quality control capabilities significantly improved the customer’s ability to develop a reliable AI model for robotic strawberry harvesting in vertical farming.

By streamlining the annotation process, ensuring precise segmentation, and saving time, Labellerr helped transform the customer’s vision into a successful AI-powered solution that enhances efficiency and productivity in vertical farming.

Need help automating your annotation and quality control processes?

Contact Labellerr today to see how we can support your AI development journey!

FAQs

1. What are the challenges of integrating AI in vertical farming?

Integrating AI in vertical farming faces challenges like high initial setup costs, the need for specialized knowledge to manage AI systems, data privacy concerns, and the complexity of collecting and processing large datasets to train AI models.

2. What are the benefits of using AI in vertical farming?

AI enhances vertical farming by optimizing resource usage (like water and nutrients), improving crop yield predictions, enabling precise environmental control, reducing labor costs through automation, and providing real-time insights for better decision-making.

3. How is AI used in vertical farming?

AI is used in vertical farming for tasks such as monitoring plant health through computer vision, automating irrigation and lighting systems, predicting optimal harvest times, and using machine learning models to adjust environmental conditions for maximum crop yield and quality.