Technology Category
Computer Vision: Medical and Biological ImagingHumans and AI: VotingSearch and Optimization: Mixed Discrete/Continuous Search
AI-Augmented Pollen Recognition in Optical and Holographic Microscopy for Veterinary Imaging
🤖 AI Summary
Pollen identification in veterinary digital in-line holographic microscopy (DIHM) images is challenging due to severe speckle noise, twin-image artifacts, and substantial domain shift from conventional bright-field imaging. Method: To address data scarcity and domain mismatch, we introduce the first optical–DIHM bimodal pollen dataset and propose three innovations: (1) spectral normalization-based Wasserstein GAN (WGAN-SN) for high-fidelity DIHM pollen image synthesis; (2) optical-image-guided DIHM bounding-box expansion; and (3) a synthetic–real hybrid training paradigm. Results: Integrated into a joint YOLOv8s detection and MobileNetV3-L classification framework, our approach improves DIHM image mAP₅₀ by 7.25 percentage points (8.15% → 15.4%) and classification accuracy by 4 percentage points (50% → 54%). Optical images achieve 91.3% mAP₅₀ and 97% accuracy. A Fréchet Inception Distance (FID) of 58.246 confirms that synthesized DIHM images are of sufficient quality to support downstream tasks.
Technology Category
Application Category
📝 Abstract
We present a comprehensive study on fully automated pollen recognition across both conventional optical and digital in-line holographic microscopy (DIHM) images of sample slides. Visually recognizing pollen in unreconstructed holographic images remains challenging due to speckle noise, twin-image artifacts and substantial divergence from bright-field appearances. We establish the performance baseline by training YOLOv8s for object detection and MobileNetV3L for classification on a dual-modality dataset of automatically annotated optical and affinely aligned DIHM images. On optical data, detection mAP50 reaches 91.3% and classification accuracy reaches 97%, whereas on DIHM data, we achieve only 8.15% for detection mAP50 and 50% for classification accuracy. Expanding the bounding boxes of pollens in DIHM images over those acquired in aligned optical images achieves 13.3% for detection mAP50 and 54% for classification accuracy. To improve object detection in DIHM images, we employ a Wasserstein GAN with spectral normalization (WGAN-SN) to create synthetic DIHM images, yielding an FID score of 58.246. Mixing real-world and synthetic data at the 1.0 : 1.5 ratio for DIHM images improves object detection up to 15.4%. These results demonstrate that GAN-based augmentation can reduce the performance divide, bringing fully automated DIHM workflows for veterinary imaging a small but important step closer to practice.
Problem
Research questions and friction points this paper is trying to address.
Automated pollen recognition in optical and holographic microscopy images
Improving detection and classification accuracy in noisy holographic data
Using GAN-based augmentation to bridge performance gaps for veterinary imaging
Innovation
Methods, ideas, or system contributions that make the work stand out.
Uses YOLOv8s and MobileNetV3L for dual-modality pollen detection and classification
Applies Wasserstein GAN with spectral normalization to generate synthetic DIHM images
Mixes real and synthetic DIHM data to improve object detection performance
🔎 Similar Papers
No similar papers found.
