From Macroscopic Recognition to Microscopic Reasoning: Advancing Plant Image Understanding

The School of EECS is hosting the following PhD Progress Review 3 Confirmation Seminar:

From Macroscopic Recognition to Microscopic Reasoning: Advancing Plant Image Understanding

Speaker: Tianqi Wei

Host: Helen Huang

Abstract:
Deep learning has substantially advanced plant image analysis, supporting applications such as plant disease recognition, disease segmentation, and plant phenotyping. However, plant images collected under practical conditions remain challenging to analyse due to complex backgrounds, large visual variations, limited annotations, and substantial differences across imaging scales. While most existing studies focus on macroscopic plant imagery, microscopic plant images introduce additional challenges that require models to recognise fine-grained biological structures and understand specialised visual content.

This thesis investigates plant image understanding across multiple scales, from macroscopic disease recognition and segmentation to microscopic image understanding. It develops datasets, benchmarks, and learning approaches for plant disease recognition in natural environments, pixel-level disease segmentation, and vision-language understanding of microscopic plant images.

Overall, this thesis aims to advance robust and generalisable plant image understanding across different visual scales and to identify the remaining limitations of modern deep learning and vision-language models in specialised plant science applications.

Bio:
Tianqi Wei is a final-year PhD candidate in Computer Science at The University of Queensland, Australia, under the supervision of Prof. Zi Huang, A/Prof. Xin Yu, and Dr. Zhi Chen. His research focuses on deep learning, computer vision, vision-language models, and their applications to plant phenotyping and agricultural image analysis. His work covers plant disease recognition and segmentation, multimodal learning, and microscopic plant image understanding.