Automated computer vision tasks are a key factor to provide high-throughput phenotyping of plants in the field, and could lead to a more efficient and sustainable agriculture. This requires analysis of data acquired in the field and development of both plant models and new computer vision methods to extract useful traits from crops, across space and time. Machine learning techniques can also help us find useful information in the data and help the farmers in their decision-making process. The roughness of in-field conditions compared to the lab’s perfectly controlled environment, as well as the diversity of crops in small market farms make for a great challenge, especially when we want to keep the costs of tools and sensors accessible to everyone. Making farmers part of these experiments is key to the approach’s success, and this is why I think it is important to develop free and open source software, and to keep the data open.