Published Paper Details:

Artificial Intelligence (AI) is transforming many fields, and plant digital phenomics is no exception. The intersection of AI and plant phenomics-- the study of plant traits and their variation-- is paving the way for significant advancements in agriculture, ecology, and plant sciences. Here's a primer on how AI is shaping this field, focusing on the journey from data to insights: Plant digital phenomics involves the collection and analysis of detailed, quantitative data on plant traits using digital tools. These traits can include growth rates, leaf shape, flower color, and stress responses. The goal is to understand the genetic, environmental, and physiological factors that influence plant development and performance. Efficient image recognition is important in crop and forest management. However, it faces many challenges, such as the large number of plant species and diseases, the variability of plant appearance, and the scarcity of labeled data for training. To address this issue, we modified a SOTA Cross-Domain Few-shot Learning (CDFSL) method based on prototypical networks and attention mechanisms. We employed attention mechanisms to perform feature extraction and prototype generation by focusing on the most relevant parts of the images, then used prototypical networks to learn the prototype of each category and classify new instances. Finally, we demonstrated the effectiveness of the modified CDFSL method on several plant and disease recognition datasets. The results showed that the modified pipeline was able to recognize several cross-domain datasets using generic representations, and achieved up to 97.85% and 95.06% classification accuracy on datasets with the same and different domains, respectively. In addition, we visualized the experimental results, demonstrating the model's stable transfer capability between datasets and the model's high visual correlation with plant and disease biological characteristics. Moreover, by extending the classes ofdifferent semantics within the training dataset, our model can be generalized to other domains, which implies broad applicability.