Deep learning techniques have been successfully applied to bioimaging problems; however, these methods are highly data demanding. An approach to deal with the lack of data and avoid overfitting is the application of data augmentation, a technique that generates new training samples from the original dataset by applying different kinds of transformations. Several tools exist to apply data augmentation in the context of image classification, but it does not exist a similar tool for the problems of localization, detection, semantic segmentation or instance segmentation that works not only with 2 dimensional images but also with multi-dimensional images (such as stacks or videos). In this paper, we present a generic strategy that can be applied to automatically augment a dataset of images, or multi-dimensional images, devoted to classification, localization, detection, semantic segmentation or instance segmentation. The augmentation method presented in this paper has been implemented in the open-source package CLoDSA. To prove the benefits of using CLoDSA, we have employed this library to improve the accuracy of models for Malaria parasite classification, stomata detection, and automatic segmentation of neural structures. CLoDSA is the first, at least up to the best of our knowledge, image augmentation library for object classification, localization, detection, semantic segmentation, and instance segmentation that works not only with 2 dimensional images but also with multi-dimensional images.

Soil nutrients and water availability are strong drivers of tropical tree species distribution across scales. However, the physiological mechanisms underlying environmental filtering along these gradients remain incompletely understood. Previous studies mostly focused on univariate variation in structural traits, but a more integrative approach combining multiple physiological traits is needed to fully portray species functional strategies. We measured nine leaf functional traits related to trees' resource capture and hydraulic strategies for 552 individuals belonging to 21 tropical tree species across an environmental gradient in Amazonian forests. Our sampling included generalist and specialist species from terra firme (TF) and seasonally flooded (SF) forests. We tested the influence of the topographic wetness index, a proxy for soil moisture and nutrient gradients, on each trait separately and on the trait integration through multivariate indices computed from the eigenvalues of a principal component analysis on the traits of the species. Finally, we evaluated intraspecific trait variability (ITV) for generalists and specialists by calculating the coefficient of variation for each trait. Results showed that (1) the environment had a greater influence on trait syndromes than single trait variation. Moreover, (2) SF specialist species expressed a stronger leaf trait coordination than TF specialist species. Furthermore, (3) the ability of generalist species to occupy a broader range of environments was not reflected by a larger ITV than specialist species but by the capacity to change trait coordination across environments. Our work highlights the need to investigate functional strategies as multidimensional syndromes in physiological trait space to fully understand and predict species distribution along environmental gradients. Read the free Plain Language Summary for this article on the Journal blog.

The increasing global population and the growing frequency of droughts shows the necessity to enhance global food production and meet future food demands. However, achieving long-term food security and effectively mitigating the impact of climate change require a critical emphasis on sustainable systems to increase food production. Hence, automatic estimation of crop production can enable breeders and farmers to make data-driven decisions to optimise resources and maximise efficiency and sustainability. In this work, we have tackled this estimation task by applying deep learning methods to images taken from a digital RGB camera. Moreover, we have improved the results of those models by feeding the models with not only images but also crop features, such as the amount of fertilisers or the amount of water. The proposed data fusion approach can be applied to convolutional- and transformer-based models obtaining good results in both cases. As a result of our work, we have produced a model that estimates crop production of wheat and spelt with an MAE of 0.666, and is a first step towards optimising resources and food production.