Colour Extraction Pipeline for Odonates using Computer Vision

2017-12-04 · Source: cs.CV updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Life Sciences & Biology, Environmental Science & Earth Systems · Depth: Expert, extended

Summary

A new computer vision pipeline has been developed to identify, segment, and extract color information from Odonates (dragonflies and damselflies) using deep neural networks. This pipeline addresses the challenge of limited annotated datasets by training on a small manually annotated set and refining it with pseudo-supervised data from citizen science platforms like GBIF. The system segments Odonates into head, thorax, abdomen, and wings, then extracts a color palette for each part. Experiments compared YOLOv11x-seg, Mask R-CNN, MaskDINO, and Mask2Former, with YOLO-exp02 achieving the best performance (mAP50 of 0.9 for bounding boxes and 0.8 for masks) after two stages of fine-tuning on a dataset of 280 images. The extracted color data, specifically mean lightness (V) values, showed a slight negative correlation with latitude and hour of day for the abdomen of *Sympetrum striolatum*.

Key takeaway

For ecologists and biodiversity researchers analyzing insect morphological traits, this pipeline offers a scalable solution to overcome manual data analysis limitations. You can leverage citizen science image data and deep learning to automate the extraction of specific body part colorations, enabling large-scale statistical analysis of ecological correlations with climate change, habitat loss, or geolocation. Consider adopting this approach to quantify and assess ecosystem biodiversity status more efficiently.

Key insights

A computer vision pipeline segments Odonates and extracts color data to enable large-scale ecological correlation analysis.

Principles

• Citizen science data can train specialized CV models.
• Pseudo-supervision refines models with limited annotations.
• YOLOv11x-seg excels in insect part segmentation.

Method

The pipeline involves manual annotation using QuPath, training and fine-tuning deep neural networks (YOLOv11x-seg, Mask R-CNN, MaskDINO, Mask2Former) on two dataset versions, and extracting color using K-Means Clustering or mean HSV values for correlation analysis.

In practice

• Use QuPath for precise biological image annotation.
• Apply K-Means Clustering for dominant color extraction.
• Analyze HSV 'V' channel for lightness correlations.

Topics

• Computer Vision
• Deep Learning
• Odonates
• Semantic Segmentation
• Instance Segmentation

Code references

• itismeganrms/colour-extraction-odonates
• andaoai/yolo-label-vs
• itismeganrms/dragonfly-formulier

Best for: AI Scientist, Computer Vision Engineer, Research Scientist

Related on AIssential

• Self-Supervised Tree-level Biomass Estimation in Urban Environments From Airborne LiDAR and Optical Observations — Self-supervised deep learning combined with LiDAR and orthophotography enables accurate, scalable urban tree biomass estimation.
• VideoSEG-O3: A Multi-turn Reinforcement Learning Framework for Reasoning Video Object Segmentation — VideoSEG-O3 introduces a multi-turn reinforcement learning framework for Reasoning Video Object Segmentation, mimicking human "coarse-to-fine" cognition.
• Using AI-Powered Behavioral Analysis to Predict Anxiety in Zebrafish — Deep learning automates precise, marker-less behavioral analysis in zebrafish, accurately predicting anxiety via pose estimation and predictive models.
• CHROMA: Detecting AI-Generated Images through Inter-Channel Color-Space Correlations — AI-generated images exhibit systematic, generator-specific inter-channel color correlation differences, useful for detection.
• Where wild things roam: Identifying wildlife with SpeciesNet — SpeciesNet is an open-source AI model for automated wildlife identification in camera trap images.
• Meta's SAM 3 (Free + Open Source) — SAM 3 enables text-prompted, open-source video object segmentation for easy analysis.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CV updates on arXiv.org.