Technology Category
Computer Vision: Object Detection & CategorizationMachine Learning: Multimodal LearningSearch and Optimization: Learning to Search
Automated Pollen Recognition in Optical and Holographic Microscopy Images
🤖 AI Summary
This study addresses the challenge of pollen grain identification in veterinary cytology by proposing a deep learning framework that fuses optical and lensless holographic microscopic images. To overcome poor hologram quality and severe label scarcity, we introduce an automated annotation pipeline coupled with bounding-box expansion to enhance model robustness on low-fidelity data. We employ YOLOv8s for multimodal object detection and MobileNetV3-Large for fine-grained classification, augmented by domain-specific data augmentation to improve generalization. On optical images, the system achieves a detection mAP₅₀ of 91.3% and classification accuracy of 97%. For holographic images—where baseline performance was extremely low (mAP₅₀ = 2.49%, accuracy = 42%)—our method elevates mAP₅₀ to 13.3% and classification accuracy to 54%. This work constitutes the first demonstration of the feasibility and efficacy of integrating low-cost lensless holographic microscopy with lightweight deep learning for veterinary pollen analysis.
Technology Category
Application Category
📝 Abstract
This study explores the application of deep learning to improve and automate pollen grain detection and classification in both optical and holographic microscopy images, with a particular focus on veterinary cytology use cases. We used YOLOv8s for object detection and MobileNetV3L for the classification task, evaluating their performance across imaging modalities. The models achieved 91.3% mAP50 for detection and 97% overall accuracy for classification on optical images, whereas the initial performance on greyscale holographic images was substantially lower. We addressed the performance gap issue through dataset expansion using automated labeling and bounding box area enlargement. These techniques, applied to holographic images, improved detection performance from 2.49% to 13.3% mAP50 and classification performance from 42% to 54%. Our work demonstrates that, at least for image classification tasks, it is possible to pair deep learning techniques with cost-effective lensless digital holographic microscopy devices.
Problem
Research questions and friction points this paper is trying to address.
Automates pollen detection and classification in microscopy images
Addresses performance gap between optical and holographic imaging modalities
Demonstrates deep learning integration with cost-effective holographic microscopy
Innovation
Methods, ideas, or system contributions that make the work stand out.
Used YOLOv8s and MobileNetV3L for detection and classification
Expanded dataset via automated labeling and bounding box enlargement
Paired deep learning with cost-effective holographic microscopy devices
🔎 Similar Papers
No similar papers found.
