Benchmarking Self-Supervised Contrastive Learning Methods for Image-Based Plant Phenotyping
TL;DR: In this paper , the authors compared two self-supervised learning methods, Momentum Contrast (MoCo) v2 and Dense Contrastive Learning (DenseCL), against the conventional supervised learning method when transferring learned representations to four downstream image-based plant phenotyping tasks: wheat head detection, plant instance detection, wheat spikelet counting and leaf counting.
read more
Abstract: The rise of self-supervised learning (SSL) methods in recent years presents an opportunity to leverage unlabeled and domain-specific datasets generated by image-based plant phenotyping platforms to accelerate plant breeding programs. Despite the surge of research on SSL, there has been a scarcity of research exploring the applications of SSL to image-based plant phenotyping tasks, particularly detection and counting tasks. We address this gap by benchmarking the performance of 2 SSL methods—momentum contrast (MoCo) v2 and dense contrastive learning (DenseCL)—against the conventional supervised learning method when transferring learned representations to 4 downstream (target) image-based plant phenotyping tasks: wheat head detection, plant instance detection, wheat spikelet counting, and leaf counting. We studied the effects of the domain of the pretraining (source) dataset on the downstream performance and the influence of redundancy in the pretraining dataset on the quality of learned representations. We also analyzed the similarity of the internal representations learned via the different pretraining methods. We find that supervised pretraining generally outperforms self-supervised pretraining and show that MoCo v2 and DenseCL learn different high-level representations compared to the supervised method. We also find that using a diverse source dataset in the same domain as or a similar domain to the target dataset maximizes performance in the downstream task. Finally, our results show that SSL methods may be more sensitive to redundancy in the pretraining dataset than the supervised pretraining method. We hope that this benchmark/evaluation study will guide practitioners in developing better SSL methods for image-based plant phenotyping.
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
Citations
Self-supervised learning advanced plant disease image classification with SimCLR
Songpol Bunyang,Natdanai Thedwichienchai,Krisna Pintong,Nuj Lael,Wuthipoom Kunaborimas,Phawit Boonrat,Thitirat Siriborvornratanakul +6 more
TL;DR: This work investigated unsupervised pre-training scenarios on unlabeled plant images across multiple architectures, including supervised fine-tuning on labeled samples, and explored the label efficiency of the self-supervised approach, acquired by fine- tuning the models on various fractions of labeled images.
4
Progress in applications of self-supervised learning to computer vision in agriculture: A systematic review
Gabriel A. Carneiro,António Cunha,Petia Radeva,Joaquim J. Sousa +3 more
A new large dataset and a transfer learning methodology for plant phenotyping in Vertical Farms
⋆. NicoSama’,Etienne David,Simone Rossetti,Alessandro Antona,Benjamin Franchetti,Fiora Pirri,Agricola Moderna,⋆. DeepPlants +7 more
- 02 Oct 2023
TL;DR: This study proposes a new plant canopy dataset, dubbed AGM of 1M images, annotated with 18 classes, an in-depth analysis of its quality for its use in transfer learning, and a methodology for detecting canopy stresses in vertical farming.
A spontaneous keypoints connection algorithm for leafy plants skeletonization and phenotypes extraction
Zhen Wang,Xiangnan He,Yu-Ting Wang,Chenxue Yang,Beilei Fan,Qingbo Zhou,Xian Li +6 more
Abstract: Introduction Leaf phenotypes are key indicators of plant growth status. Existing deep learning–based leaf skeletonization typically requires extensive manual labeling, long training, and predefined keypoints, which limits scalability. We developed a training-free and label-free approach that connects spontaneously detected keypoints to generate leaf skeletons for leafy plants. Methods The method comprises random seed-point generation and adaptive keypoint connection. For plants with random leaf morphology, we determine a threshold for the angle difference among any three consecutive adjacent points and iteratively identify keypoints within circular search neighborhoods to trace leaf skeletons. For plants with regular leaf morphology, we fit the skeleton trajectory by minimizing curvature. We validated the approach on vertical and front-view images of orchids (covering random and regular morphological cases) and extracted five phenotypic parameters from the resulting skeletons. Generalization was further assessed on a maize image dataset. Results On orchid images, the proposed approach achieved an average curvature fitting error of 0.12 and an average leaf recall of 92%. Five orchid phenotypic parameters were accurately derived from the skeletons. The method also showed effective skeletonization on maize, indicating cross-species applicability. Discussion By eliminating manual labels and training, this approach reduces annotation effort and computational overhead while enabling precise geometric phenotype calculation from skeleton-based keypoints. Its effectiveness on both randomly distributed and regularly shaped leafy plants suggests suitability for high-throughput plant phenotyping workflows.
References
Deep Residual Learning for Image Recognition
Kaiming He,Xiangyu Zhang,Shaoqing Ren,Jian Sun +3 more
- 27 Jun 2016
TL;DR: In this article, the authors proposed a residual learning framework to ease the training of networks that are substantially deeper than those used previously, which won the 1st place on the ILSVRC 2015 classification task.
ImageNet: A large-scale hierarchical image database
Jia Deng,Wei Dong,Richard Socher,Li-Jia Li,Kai Li,Li Fei-Fei +5 more
- 20 Jun 2009
TL;DR: A new database called “ImageNet” is introduced, a large-scale ontology of images built upon the backbone of the WordNet structure, much larger in scale and diversity and much more accurate than the current image datasets.
Deep learning
TL;DR: Deep learning is making major advances in solving problems that have resisted the best attempts of the artificial intelligence community for many years, and will have many more successes in the near future because it requires very little engineering by hand and can easily take advantage of increases in the amount of available computation and data.
67K
Microsoft COCO: Common Objects in Context
Tsung-Yi Lin,Michael Maire,Serge Belongie,James Hays,Pietro Perona,Deva Ramanan,Piotr Dollár,C. Lawrence Zitnick +7 more
- 06 Sep 2014
TL;DR: A new dataset with the goal of advancing the state-of-the-art in object recognition by placing the question of object recognition in the context of the broader question of scene understanding by gathering images of complex everyday scenes containing common objects in their natural context.
Fully convolutional networks for semantic segmentation
Jonathan Long,Evan Shelhamer,Trevor Darrell +2 more
- 07 Jun 2015
TL;DR: The key insight is to build “fully convolutional” networks that take input of arbitrary size and produce correspondingly-sized output with efficient inference and learning.