ICPR 2024 LEAF INSPECT
Introduction
The advancement in imaging with computer vision based methods have facilitated non-invasive plant trait quantification, a key component of Precision Agriculture. The majority of plant traits are primarily derived from leaf-level analysis of plant images, underlining the importance of leaf instance segmentation tasks in Precision Agriculture. In this challenge, we propose a novel plant image dataset with leaf-level annotations to facilitate the development of accurate computer vision based methods for leaf instance segmentation.
Challenge Description
Recent datasets comprise of model plants (https: //plant-image-analysis.org) primarily Arabidopsis thaliana, maize and sorghum because of their brief life cycle and compact dimensions. In contrast, classical plants such as rice, and wheat exhibit high inter-class and intra-class variances with respect to leaf's shape, size etc. Thus, current computer vision based methods do not compensate for the research on classical plants with complex growth patterns .
We see this as an opportunity and propose a large classical plant dataset with leaf-level annotations to bridge the aforementioned plant species gap. Unlike common target objects such as table, chair etc. leaf exhibits stochasticity with respect to its shapes, structures and appearances corresponding to different plant species. The novel dataset in the challenge along with the algorithms will facilitate computer vision research on non-rigid objects with high degree of self-similarity and self-occlusions.
Sample Images of Rice at different growth stages
Important Dates
Apr 26 2024 Registration Open
May 9 2024 Training Data Online
June 9 2024 Test data ready for download
Jul 19 2024 Challenge Submission Deadline
Jul 28 2024 Results Announcement
Aug 18 2024 Challenge Reports Deadline
Sep 2 2024 Camera-Ready Papers Due