🌿 Leaf Segmentation for Plant Phenotyping and Medicinal Plant Identification
In collaboration with colleagues at Gauhati University, I have contributed to research focused on automating leaf segmentation—a crucial step in plant phenotyping and medicinal plant identification. This work aims to develop systems capable of accurately isolating individual leaves from complex backgrounds, facilitating detailed analysis of plant traits such as leaf count, area, and morphology. Such analyses are vital for understanding plant health, growth patterns, and for applications in agriculture and botanical studies.
Examples of some segmented leaves.
From a technical perspective, our approach leverages both classical image processing techniques and advanced deep learning models. Notably, we have explored the use of transformer-based architectures for image segmentation tasks. Transformers, known for their ability to capture long-range dependencies in data, offer advantages over traditional convolutional neural networks (CNNs) by providing more contextual understanding of images. This shift towards transformer architectures addresses limitations in capturing complex spatial relationships and varying object scales in plant imagery (Chetia et al., 2025).
In the realm of medicinal plant identification, our team developed a custom CNN model trained on a self-curated dataset comprising over 42,000 images of 50 medicinal plant species native to Assam, India. This model achieved high accuracy rates across multiple datasets, demonstrating its effectiveness in distinguishing visually similar plant species—a task traditionally reliant on expert human intervention (Chetia et al., 2024).
These collaborative efforts underscore the potential of integrating advanced machine learning techniques with botanical research. By automating the segmentation and identification processes, we aim to enhance the efficiency of plant phenotyping and support the preservation and utilization of medicinal plant knowledge.
References
2025
arXiv
Image Segmentation with transformers: An Overview, Challenges and Future
Image segmentation, a key task in computer vision, has traditionally relied on convolutional neural networks (CNNs), yet these models struggle with capturing complex spatial dependencies, objects with varying scales, need for manually crafted architecture components and contextual information. This paper explores the shortcomings of CNN-based models and the shift towards transformer architectures -to overcome those limitations. This work reviews state-of-the-art transformer-based segmentation models, addressing segmentation-specific challenges and their solutions. The paper discusses current challenges in transformer-based segmentation and outlines promising future trends, such as lightweight architectures and enhanced data efficiency. This survey serves as a guide for understanding the impact of transformers in advancing segmentation capabilities and overcoming the limitations of traditional models.
@article{chetia2025segtrans,title={Image Segmentation with transformers: An Overview, Challenges and Future},author={Chetia, Deepjyoti and Dutta, Debasish and Kalita, Sanjib Kr},journal={arXiv preprint arXiv:2501.09372},year={2025},archiveprefix={arXiv},primaryclass={cs.CV},dimensions={true},keywords={arxiv, Image Segmentation, Transformer, Vision Transformers (ViTs), Self-Attention Mechanism, Deep Learning},doi={https://doi.org/10.48550/arXiv.2501.09372},url={https://arxiv.org/abs/2501.09372},}
2024
Springer
Identification of Traditional Medicinal Plant Leaves Using an effective Deep Learning model and Self-Curated Dataset
Deepjyoti Chetia, Sanjib Kr Kalita, Prof Partha Pratim Baruah, and 2 more authors
Medicinal plants have been a key component in producing traditional and modern medicines, especially in the field of Ayurveda, an ancient Indian medical system. Producing these medicines and collecting and extracting the right plant is a crucial step due to the visually similar nature of some plants. The extraction of these plants from nonmedicinal plants requires human expert intervention. To solve the issue of accurate plant identification and reduce the need for a human expert in the collection process; employing computer vision methods will be efficient and beneficial. In this paper, we have proposed a model that solves such issues. The proposed model is a custom convolutional neural network (CNN) architecture with 6 convolution layers, max-pooling layers, and dense layers. The model was tested on three different datasets named Indian Medicinal Leaves Image Dataset,MED117 Medicinal Plant Leaf Dataset, and the self-curated dataset by the authors. The proposed model achieved respective accuracies of 99.5%, 98.4%, and 99.7% using various optimizers including Adam, RMSprop, and SGD with momentum.
@inproceedings{chetia2025med,title={Identification of Traditional Medicinal Plant Leaves Using an effective Deep Learning model and Self-Curated Dataset},author={Chetia, Deepjyoti and Kalita, Sanjib Kr and Baruah, Prof Partha Pratim and Dutta, Debasish and Akhter, Tanaz},journal={International Conference on Advanced Network Technologies and Intelligent Computing},year={2024},pages={342--356},archiveprefix={arXiv},dimensions={true},keywords={arxiv, Medicinal plant identification · Medicinal plant leaf dataset deep learning · Convolutional Neural Network(CNN)},doi={https://doi.org/10.1007/978-3-031-83793-7_22},url={https://arxiv.org/abs/2501.09363},}
