Transforming Syngenta’s lab inspection with Vision AI

Syngenta's integration of AI, powered by Google Cloud's Vertex AI and developed in collaboration with ML6, has revolutionized their crop protection research. By automating the analysis of half a million wells annually, the solution has doubled analysis speed while significantly enhancing scoring reliability. This transformation addresses the challenges of subjective manual scoring and resource constraints, freeing lab experts to focus on higher-level research. The AI's ability to accurately detect and interpret plant disease symptoms, coupled with custom business logic, ensures consistent and interpretable results, accelerating the development of more effective and sustainable agricultural solutions. Ultimately, this innovation contributes to safeguarding global food security by enabling faster discovery of crucial crop protection products, while minimizing environmental impact.

‍

Syngenta is a leading science-based agriculture tech company that specialises in producing seeds and pesticides. They are accelerating innovation by using AI at the centre of their research and development programs. AI is already changing the way medicines are discovered. Machine learning and other technologies are making the hunt for new pharmaceuticals quicker, cheaper and more effective. Syngenta’s goal is to do the same for agriculture – to discover new, more effective crop protection solutions that safeguard the world’s food against diseases, weeds and pests, while also protecting ecosystems.

‍

At Syngenta, lab researchers conduct manual inspections of wells containing treated leaves with different amounts and combinations of fungicides. Each well is assigned an efficacy score of 0, 20, 50, 70, 90, or 100% based on the effectiveness of the fungicide and thus the health of the plant. However, this process is subjective—what one researcher might rate as 50%, another might rate as 70%. This variability in scoring leads to inconsistencies in the evaluation of fungicide efficacy. Additionally, the process is time-intensive and relies on the expertise of skilled lab researchers, a resource that is becoming increasingly difficult to source. Syngenta needed a solution to enhance the consistency of the rating process while simultaneously augmenting efficiency, thereby enabling lab experts to concentrate on priority tasks.

Lab researchers manually inspect and assign efficacy scores to wells containing treated leaves. This process is time consuming, can be subjective, leading to inconsistencies and inefficiencies. Furthermore, staffing skilled lab researchers accordingly became more difficult. Syngenta needed a solution to improve consistency, increase efficiency, and allow lab experts to focus on higher priority tasks.

‍

ML6's expertise in AI and its proven track record in developing state-of-the-art solutions combined with Google Cloud’s leading AI products created an ideal setup for Syngenta: The teams collaborated closely with Syngenta to develop a tailored vision AI solution to support the lab experts in this process, leveraging Vertex AI. ML6 developed a semantic segmentation model to detect the unhealthy parts of the leaves. It is able to recognize different symptoms of plant diseases ranging from brown spotting to white discoloration. This is combined with custom business logic developed and tested in co-creation with the domain experts from Syngenta. The output of the vision AI system gives an efficacy score of the pesticide for each well. The AI prediction then supports the lab expert to make the final decision on the efficacy scores.

‍

A semantic segmentation model was created to detect unhealthy areas on leaves and recognize various disease symptoms with a vision AI solution using Vertex AI: Combined with custom business logic, jointly developed with Syngenta’s domain experts, the system provides an efficacy score for each pesticide well, aiding lab experts in making final decisions.

‍

Syngenta processes approximately half a million wells annually. The implemented solution has doubled the analysis speed while ensuring more reliable scoring. The integration of AI with tailored business intelligence logic has been instrumental in project success, offering easily interpretable results verified by lab experts offering them to be more accurate and to perform research more efficiently.