An artificial intelligence tool developed by AGROSAVIA will help demonstrate the quality of Colombian cocoa to the world

• Researchers at AGROSAVIA have developed IA4CACAO, an artificial intelligence-based tool that automatically evaluates the fermentation level of cocoa beans. The technology aims to make cocoa quality classification more objective, strengthen confidence among buyers and producers, and enhance the competitiveness of Colombian cocoa in both domestic and international markets.

 

Mosquera, Cundinamarca. July 8, 2026. Cocoa quality begins long before it is transformed into chocolate. One of the processes that most significantly influences its aroma, flavor, and commercial value is fermentation, a stage that has traditionally relied on the expertise of evaluators who assess cocoa beans through the cut test. However, differences in interpretation among evaluators can lead to inconsistencies in quality classification.

To provide greater accuracy and objectivity in this process, researchers from the Corporación colombiana de investigación agropecuaria (AGROSAVIA) developed IA4CACAO, an artificial intelligence-based tool that automatically classifies the fermentation level of cocoa beans using images obtained during the cut test.

The findings of this research were published in the scientific article IA4CACAO: Deep Learning-Based Classification of Cocoa Bean Fermentation Levels Using Cut Test Images in Colombia, demonstrating the potential of this technology to support the quality assessment of Colombian cocoa.

 

A Tool That Learns to Recognize Cocoa Quality

IA4CACAO uses deep learning models trained on thousands of images of cocoa beans produced in different regions of Colombia. Through this training process, the system identifies visual patterns associated with different fermentation levels and performs automatic classification with a high degree of accuracy.

"Artificial intelligence does not replace the expertise of the evaluator; rather, it complements it by providing consistent results and reducing the variability inherent in visual inspection," explains Dr. César Vargas, Researcher at the Tibaitatá Research and Technology Transfer Center and Lead Researcher of the study.

Unlike other applications developed around the world, IA4CACAO was designed using data generated from Colombian cocoa materials, improving its ability to address the specific characteristics of the country's cocoa production.

 

Greater Confidence for Producers, Buyers, and Exporters

Fermentation is one of the primary criteria used to determine cocoa quality and its commercial value. Having a tool that standardizes this evaluation represents an opportunity to strengthen trust among producers, associations, traders, and the cocoa industry.

A more objective classification facilitates the formation of homogeneous lots, supports traceability processes, and helps demonstrate the quality of Colombian cocoa—an increasingly important attribute in specialty markets.

In addition, the technology can support technical assistance and training programs by enabling producers and extension specialists to more accurately assess post-harvest processing results and make informed decisions to improve product quality.

 

From the Laboratory to the Field

Although IA4CACAO is currently in the research and validation stage, AGROSAVIA's goal is to further develop this technology into an accessible tool for the productive sector.

Researchers envision its implementation through mobile applications, tablets, or portable devices that allow evaluations to be conducted directly on farms, collection centers, or post-harvest processing facilities, without requiring specialized infrastructure.

Future developments also include expanding the image database with samples from different cocoa varieties and regions across the country, incorporating new features for automated batch analysis, and strengthening integration with traceability and quality certification systems.

 

Artificial Intelligence at the Service of Colombian Agriculture

The development of IA4CACAO demonstrates how digital technologies can become valuable allies for agriculture—not by replacing producers' expertise, but by strengthening it with tools that support better decision-making.

By integrating artificial intelligence, computer vision, and agronomic knowledge, AGROSAVIA continues to develop innovative solutions that enhance the competitiveness of Colombian cocoa, reinforce confidence in its quality, and bring science closer to the real needs of the productive sector.

Click in the following link to access the scientific article: IA4CACAO: Deep Learning-Based Classification of Cocoa Bean Fermentation Levels Using Cut Test Images in Colombia