The application of methodologies to measure specific organism’s (e.g. plant, insect etc.) traits (morphology, growth etc.) related to its structure and function is termed as phenotyping. With the emergence of low-cost and high-resolution multi-modal cameras, acquisition of 2D and 3D data permits high-throughput micro and macro analysis. This is a rapidly growing field at the interface of biology and computer vision (CV) termed-High-throughput Vision based Phenotyping. Agriculture Automation & Information systems nowadays extensively use high speed image sensors along with multi-scale and hyperspectral imaging methods to effectively capture plant traits, facilitating comprehensive monitoring and analysis. Vision-based plant phenotyping encounters difficulties in handling large and intricate data, coping with unpredictable environmental changes, and combining different types of information. Overcoming these challenges requires advanced methods to make sense of the data and help improve crops and farming practices effectively.

In contrast to common datasets like ImageNet, MSCOCO, PASCAL VOC datasets, organisms present unique challenges due to their dynamic nature and inherent variability. These challenges include tracking deformable objects like microbes in microscopy, multi-label segmentation of self-similar objects such as leaves in plants and fish, and 3D reconstruction in complex environments with overlapping surfaces, such as plant structures. Natural conditions, such as those found in agricultural fields, greenhouses, forests, and marine ecosystems, further complicate image analysis. This workshop will spotlight the challenges of applying computer vision to plant/animal phenotyping and agricultural research, aiming to showcase advanced methodologies, identify critical unresolved issues, and engage computer scientists interested in this domain. As effective plant phenotyping is crucial for sustainability, fostering community involvement and welcoming computer vision experts into this field is of paramount importance.

Poster Showcase Topics (not limited to):

● Farmland pattern classification, detection, and segmentation from agricultural/phenotyping imagery.

● Resources and dataset benchmarks for agricultural imagery based pattern analysis.

● Data fusion of multi/hyper-spectral image data and multi-modal data sources

● Self, semi, and weakly supervised methods for agricultural/phenotyping imagery

● Transfer learning and domain adaptation

● Generative AI for plant/animal/microorganisms phenotyping

● 3D modeling and segmentation, UAV based field phenotyping

● Efficient data sampling methods and learning with limited training data or in presence of noisy, sparse, and imbalanced annotations.

● Computer vision applications which promote the study or adoption of sustainable agriculture

We invite posters from already published research papers in the area of plant/animal phenotyping or computer vision for agriculture or posters on some current ongoing work that present innovative ideas, methods, and applications in the area of phenotyping using computer vision. We encourage submissions that have interdisciplinary collaborations between machine learning/computer vision and problem domain experts. We especially encourage work where machine learning and in particular representation learning could meaningfully amplify existing efforts for the phenotyping area.

Registration Link for Poster showcase: https://forms.gle/gH152Zb6vBVGMquo9 

1. This workshop aims to provide a forum to both show current relevant efforts in interdisciplinary areas between computer vision and agriculture, and to encourage further research and conversations within the computer vision community to tackle impactful agriculture-vision problems.

2. Open new avenues for collaboration amongst researchers and agricultural organizations, NGOs, local government bodies and other organizations to enable CV based beneficial research.

3. Define new, quantifiable, and impactful research questions/areas for the CV community.

4. New tools, datasets and projects can be promoted which would interest CV researchers to apply their skills on challenging phenomics problems.