Researchers, students, agricultural technicians and farmers can now count on a database containing images with the accurate phytopathological description of the main symptoms and signs of diseases in several crops. The database named Digipathos is free and available for public access online. There are almost 3,000 digital photographs of the main cash crops such as soybeans, coffee, rice, beans, wheat, maize and fruit species, among others (see box below), which can be consulted and downloaded, especially for use in technical, academic and research studies. The repository has already been used by the international scientific community in Africa, China and India, besides Brazil. This catalogue of images of diseases that attack the plant species is extremely relevant to facilitate early diagnosis. The action is fundamental to ensure food security and prevent losses, but the constant monitoring of the plants in the field becomes impracticable depending on the size of the cropland and on human ability to detect the diseases. Catalogued crop types Phytopathologists of 14 Embrapa research centers spread throughout the country have collaborated to the initiative, feeding into the repository. Besides soybeans, coffee, rice, beans, wheat and maize, the catalogue includes: cotton, sugarcane, sorghum, citrus, grapevines, pineapple, cupuaçu, açaí, anthurium, melon, oil palm, coconut palm and black pepper. “A database with illustrative images of plant diseases especially helps the professionals involved with agricultural production, as they often come across plant health problems in croplands whose diagnosis is difficult or generates doubts", asserts Flávia Rodrigues Patrício, a researcher from Instituto Biológico (IB), located in the state of São Paulo. "The correct diagnosis is fundamental so that there are right decisions concerning control and management measures", she adds. Similar diseases can be caused by different pathogens. For example, in coffee crops, branches can dry out either due to leafspot, a disease caused by a fungus named Phoma tarda, or halo blight, a disease caused by a bacterium, Pseudomonas syringae pv. garcae, or even due to abiotic factors, such as excess load and nutritional deficiencies. "If there is a misdiagnosis, the correct measures will not be applied in time and growers could endure considerable losses", the researcher details. Automated diagnosis The database was also created to become a reference for the development of methods to detect and automatically recognize plant diseases. The idea is to expand it to include symptoms and detailed descriptions of the causes and consequences of each disease. "All the images were labelled by experienced phytopathologists, thus supplying reliable data to train the algorithms developed", explains the researcher from Embrapa Agricultural Informatics Jayme Barbedo, who coordinates Digipathos. An app to diagnose diseases At Embrapa, research with digital processing aims to develop technologies that support the automated diagnosis of plants of commercial and social interest in Brazil. Therefore, the works in progress have also focused in the development of methods to generate reliable, computer-run diagnoses based on images supplied by users. The team aims to create an app and a web service to help farmers to directly identify diseases that are attacking the crops in the fields. The first version of this technology is due to be tested in the first half of 2019. The researchers Bernardo Halfeld, Kátia Nechet and Daniel Terao, from Embrapa Environment, participated in the process of feeding the bank and agree that the differential of this type of technology is the large variability of symptoms it contains, which translates into higher precision in the final diagnosis. Nechet observes that the signals and symptoms of plant diseases that were previously only described in books, often without illustrations, hindered understanding and a precise diagnosis. "Now, through digital images, the identification of phytopathological problems in the field will be facilitated with greater speed. This quick diagnosis will contribute to reduce losses in the production sector", she concludes. The researcher Bernardo Halfeld explains that the tool was thought to directly help in the recognition of specific disease patterns, is caused by biotic factors, that is, by the action of microorganisms (for instance: virus, fungi, bacteria) or by abiotic factors, which are caused by phytotoxicity or by the influence of elements from the environment, such as solar radiation, temperature, nutritional factors, among others. "The database is going to facilitate the work of research, as it uses verified images, determining with higher precision the causal agent and the best approach to be adopted. It is a tool that operates at all levels and, in practice, it will inform the determination of the best kind of management, reducing chemical applications in the crops, reducing production costs, improving disease control and productivity, as well as supporting the scientific studies on the subject", Halfeld assesses. The database is the result of a partnership between Embrapa and Instituto Biológico, with financial support from the São Paulo State Research Support Foundation (Fapesp). It is free and open, as long as the rules of publication and citation are followed. Commercial use is not allowed unless with Embrapa's express authorization. To reference the database, it is necessary to cite the authors, according to the terms and conditions available in the repository. Practical application in research Instituto Biológico develops research involving diagnosis techniques, phytopathogen populations, cultivar resistance, disease management, biological control and chemical control of plant diseases, among others. The database informs the correct identification of the diseases to be studied, for researchers, undergraduate and graduate students involved in research, agricultural technicians, agronomists and external professionals who also collaborate in the studies. The repository can also be an important tool for another activity provided by Instituto Biológico: lab exams for the diagnosis of more complex pests and diseases that are hard to diagnose through their visual symptoms, reducing farmers' doubts and the need to submit materials for analysis. Research with coffee crops that can benefit from this material includes the study that correlates the incidence of pests and diseases with the climate, conducted in different coffee-producing cities in São Paulo state, in collaboration with the Campinas Agronomic Institute (IAC), the Food Technology Institute (Ital) and Embrapa Coffee. The work aims to guide the growers and extension agents with regard to the regions and times of the year that are most favorable for each disease, and to detecting possible influences of pests and diseases on the quality of the coffee drink produced in the state of São Paulo. "In this study, external partners have assisted in the survey of the diseases, and the material available in the database can help clearing the doubts that often occur in field conditions. The database can also help in the training of graduate and undergraduate students involved in studies on the biological control, chemical control and management of plant diseases in field and greenhouse conditions", Flávia Patrício says. The graduate student Juliana Mariana Macedo Araújo, who takes a Master's course in Electric Engineering, uses the content in a research project on signal processing and analysis from the Pontifical Catholic University of Minas Gerais (PUC Minas). The goal is to create a system for the automatic detection of soybean diseases based on images of the plant leaf, with analysis of the color, texture and format of the injured region. "The database has been helping me a lot, as I did not find an open database with this number of images available", she reports. Students create new technologies with Digipathos The students Sara Alves de Godoi and Pedro Luiz Goulart de Melo, from the higher degree course in Systems Technology for the Internet at Faculdade de Tecnologia (Fatec) in São Roque, SP, used the repository in their final dissertation. "The intention was to help farmers in the São Roque region detect diseases in fruits and vegetables. Sara elaborated an app for mobile devices in which farmers capture the images and send them to a back-end system. Pedro elaborated this system so that an agronomist receives the images and replies to the farmer", reports the course lecturer Fernando Di Gianni. He recalls that the students had the database structure, however, they needed real content, such as images and remaining information on the diseases. "That was when we made contact with Embrapa to request access to the image repository, so that agronomists can compare the images received with the ones from Digipathos, facilitating the identification of crop diseases", the professor states. "With this repository, we can stimulate other students to continue with the work that was carried out and improve the analysis of the images with artificial intelligence techniques, for example", he anticipates. Two other studies developed at Fatec have focused on the monitoring of hydroponic crops and will give continuity to the development of an app for mobile devices that collects data from the back-end system and informs the farmer if there is an anomaly in the crop. Moreover, the college encourages perfecting the studies using machine learning. "The interaction with Embrapa is very productive by allowing students to expand the knowledge base, collaborating to solving the most diverse problems found in farmers' daily routine", analyzes Di Gianni. Photo: Marcos Vicente Translation: Mariana Medeiros