Identifying Planetary Nebulae Candidates from IPHAS and VPHAS+ Surveys via YOLO Deep Learning Models
Abstract
We present a deep learning detection method capable of effectively identifying planetary nebulae (PNe) on IPHAS and VPHAS+ surveys. Our detection pipeline is based on the YOLO architecture and is adaptable to other datasets. Using the HiPS system, we preprocess and augment the r, i, and Hα bands to produce two red, green, and blue (RGB) image datasets: (1) 640 × 640 px images at full resolution, and (2) 1 280 × 1 280 px images designed to capture extended objects up to $0.2$ deg. Images of ∼ 2600 PNe are labelled using the HASH database. The loss function and evaluation metrics are adapted to the diffuse and low-contrast nature of PNe; a very low intersection over union (IoU) threshold (1%) is used. A k-fold cross validation shows that the YOLOv12m model achieves an average precision of $0.979$ on the first dataset. At a confidence threshold of $0.15$, it achieves a precision rate of 98.7% and a recall rate of 97.4%. We deployed our system (on 1280 × 1280 images) on the full surveys' coverage (∼ 3755 deg2), and present a preliminary analysis of the detection results. The system is robust against artifacts such as CCD corner defects and HiPS tile joints. Our convolutional-neural-network-based model achieved excellent results on the test set, comparable to-or exceeding-that of a recent Swin Transformer approach. However, a challenge persists when applying deep learning models to extensive and densely populated Galactic fields: the development of an automated system that can effectively discriminate PNe from mimics that human experts are able to distinguish (e.g., H II regions). We outline possible improvements for future work.
Versions
Related
- Classification of Galactic Planetary Nebulae Among HASH Candidates
- X-Shooter view of planetary nebulae microstructures. I. The case of NGC 7009 knots
- Proto-Planetary Nebulae Are Not the Precursors of Planetary Nebulae With Close Binary Central Stars
- Newly Identified Planetary Nebula Candidates in M31 using Deep-Learning Search of Hubble Space Telescope Images
- The jet-shaped pipe morphology in planetary nebulae and core-collapse supernova remnants
BibTeX
@misc{serandour2026identifying,
title = {{Identifying Planetary Nebulae Candidates from IPHAS and VPHAS+ Surveys via YOLO Deep Learning Models}},
author = {Titouan Sérandour and Anne-Laure Méalier and Bruno Colicchio and Dominique Daniel and Philippe Gilberton and Philippe Oros and Thomas Boch and Djaffar Ould Abdeslam},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.19760},
eprint = {2610.19760},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.19760},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Identifying Planetary Nebulae Candidates from IPHAS and VPHAS+ Surveys via YOLO Deep Learning Models
AU - Titouan Sérandour
AU - Anne-Laure Méalier
AU - Bruno Colicchio
AU - Dominique Daniel
AU - Philippe Gilberton
AU - Philippe Oros
AU - Thomas Boch
AU - Djaffar Ould Abdeslam
PY - 2026
DA - 2026/10/14
PB - arXiv
DO - 10.48550/arxiv.2610.19760
UR - https://arxiv.org/abs/2610.19760
AB - We present a deep learning detection method capable of effectively identifying planetary nebulae (PNe) on IPHAS and VPHAS+ surveys. Our detection pipeline is based on the YOLO architecture and is adaptable to other datasets. Using the HiPS system, we preprocess and augment the r, i, and Hα bands to produce two red, green, and blue (RGB) image datasets: (1) 640 × 640 px images at full resolution, and (2) 1 280 × 1 280 px images designed to capture extended objects up to $0.2$ deg. Images of ∼ 2600 PNe are labelled using the HASH database. The loss function and evaluation metrics are adapted to the diffuse and low-contrast nature of PNe; a very low intersection over union (IoU) threshold (1%) is used. A k-fold cross validation shows that the YOLOv12m model achieves an average precision of $0.979$ on the first dataset. At a confidence threshold of $0.15$, it achieves a precision rate of 98.7% and a recall rate of 97.4%. We deployed our system (on 1280 × 1280 images) on the full surveys' coverage (∼ 3755 deg^2), and present a preliminary analysis of the detection results. The system is robust against artifacts such as CCD corner defects and HiPS tile joints. Our convolutional-neural-network-based model achieved excellent results on the test set, comparable to-or exceeding-that of a recent Swin Transformer approach. However, a challenge persists when applying deep learning models to extensive and densely populated Galactic fields: the development of an automated system that can effectively discriminate PNe from mimics that human experts are able to distinguish (e.g., H II regions). We outline possible improvements for future work.
ER -