Agricultural computer vision tasks are highly diverse, consisting of varied sensor inputs and many potential perception output tasks. Moreover, thousands of crop types exist across heterogeneous farming conditions globally, introducing substantial variability in potential sensor environments. To overcome such massive variability in the face of limited data, there is a need to develop and apply new methods for data efficiency and generalizability for agricultural computer vision. In this workshop, we seek presentation and paper submissions that aim to improve data efficiency and generalizability via novel techniques in multi-modal sensor fusion, domain adaptation, semi-supervised learning, learnable augmentations, transfer learning, and synthetic data generation, among other emerging techniques. The DEGA-CV workshop encourages presentation and paper submissions to take advantage of more than 25 centralized and standardized agricultural datasets available via the AgML Python API, which will also be used as part of an ”Data Efficiency and Generalizability” challenge with multiple prizes offered.