On Monday, June 19th, the Institute of Communication and Computer Systems (ICCS) participated in the 4th International Workshop on Agriculture-Vision held alongside the esteemed IEEE/CVF Computer Vision and Pattern Recognition Conference 2023 (CVPR). The conference is one of the premier ones concerning computer vision and pattern recognition. Additionally, its workshops are known to be prestigious meeting places for researchers in the specific field. More specifically, regarding agriculture-related computer vision research, CVPR and its associated workshops are particularly crucial since there is no focused agricultural conference for this domain.
Therefore, it was a splendid opportunity for ICCS to contribute to the workshop with two papers, “Mushroom Segmentation and 3D Pose Estimation from Point Clouds using Fully Convolutional Geometric Features and Implicit Pose Encoding” and the “Realistic Synthetic Mushroom Scenes Dataset”. George Retsinas, Niki Efthymiou, and Petros Maragos authored the first one. It showcases an innovative approach to address the challenges of mushroom segmentation and 3D pose estimation using point cloud data. Our method leverages deep learning techniques based on novel pose embeddings and a synthetic dataset of mushroom scenes. Our research demonstrates the potential of this approach to advance agricultural computer vision.
The second paper included contributions by Dafni Anagnostopoulou, George Retsinas, Niki Efthymiou, Panagiotis Fiintisis, Panagiotis Filntisis and Petros Maragos. It features 15,000 high-quality images of mushrooms in diverse settings, complemented by detailed annotations. This dataset is a valuable resource for tackling critical issues in mushroom detection, instance segmentation, and 3D pose estimation, especially in the context of automating labour-intensive tasks in mushroom harvesting. Additionally, the presentation included a flexible three-step pipeline for generating annotated synthetic images, which has promising applications in agricultural computer vision research.
Congratulations to all the authors for helping Project SoftGrip travel across continents once more and disseminating its results to the international research community.
