Toward a Public Dataset of Wide-Field Astronomical Images Captured

with Smartphones

Olivier Parisot

and Diogo Ramalho Fernandes

Luxembourg Institute of Science and Technology (LIST),

5 Avenue des Hauts-Fourneaux,

4362 Esch-sur-Alzette, Luxembourg

Keywords:

Smartphone, Astronomy, Deep Learning.

Abstract:

Smartphone photography has improved considerably over the years, particularly in low-light conditions.

Thanks to better sensors, advanced noise reduction and night modes, smartphones can now capture detailed

wide-ﬁeld astronomical images, from the Moon to deep sky objects. These innovations have transformed them

into pocket telescopes, making astrophotography more accessible than ever. In this paper, we present AstroS-

martphoneDataset, a dataset of wide-ﬁeld astronomical images collected since 2021 with various smartphones,

and we show how we have used these images to train Deep Learning models for trails detection.

1 INTRODUCTION

In recent years, astrophotography has become in-

creasingly accessible, whereas it was once limited to

experienced enthusiasts with specialized equipment

(Woodhouse, 2015). Thus, there are many ways to

take pictures of celestial objects such as the Sun, the

Moon, the planets of the Solar System and Deep Sky

Objects (DSO) – for both amateur and professional

astronomers.

Installations combining small optical instruments

(such as 60mm diameter refractors) with inexpen-

sive CMOS cameras enable astrophotography and

Electronically Assisted Astronomy (EAA, sometimes

called Video Astronomy) to image most DSO with

good details (Ashley, 2017), but the process requires

speciﬁc equipment and advanced technical skill to set

it up and use it properly. To obtain high-quality re-

sults, large-diameter instruments coupled with ultra-

sensitive CMOS cameras enable displaying stunning

images of Jupiter, exploring outer Solar System (Ari-

matsu, 2025), discovering unknown DSO (Drechsler

et al., 2023), or even detecting transient objects (Ro-

manov, 2022). Here again, more expensive and com-

plex equipment is required, as well as considerable

know-how.

Recently, modern smart telescopes have taken a

https://orcid.org/0000-0002-3293-3628

https://orcid.org/0009-0008-3187-3468

major step forward by automating all the tedious tasks

involved, making it possible for the general public to

observe the sky in a comfortable way (Parisot et al.,

2022). As well as being fun and educational, these

automated instruments enable astronomers to capture

large quantities of data, which is essential for produc-

ing quality results (in particular to optimise the signal-

to-noise ratio of the images obtained).

In recent years, smartphones have achieved re-

markable progress in photography (Fang et al., 2023),

both in sensor technology and onboard software pro-

cessing, matching or even exceeding the performance

of digital cameras. It is therefore possible to use

smartphones as imagers, whether or not in combina-

tion with complementary devices. One example is

Hestia, an instrument that ampliﬁes the capabilities

of a smartphone to photograph the Moon, the Sun and

even the brightest DSO

In this work, we present AstroSmartphoneDataset,

a dataset of images captured between 2021 and 2025

with four different smartphones, using the device’s

built-in sensor and processing, without any additional

optical instruments. Such dataset has both academic

and practical applications. Scientiﬁcally, it enables to

design automated systems for satellite detection and

space debris monitoring (crucial for Space Situational

Awareness), by serving as valuable training data for

Deep Learning models in automated object recogni-

https://vaonis.com/fr/pages/product/hestia

644

Parisot, O., Fernandes and D. R.

Toward a Public Dataset of Wide-Field Astronomical Images Captured with Smartphones.

DOI: 10.5220/0013642300003967

In Proceedings of the 14th International Conference on Data Science, Technology and Applications (DATA 2025), pages 644-650

ISBN: 978-989-758-758-0; ISSN: 2184-285X

tion and classiﬁcation. Moreover, it can be used as

educational resources for teaching astronomy and im-

age processing techniques.

The present paper is structured as follows. In Sec-

tion 2, we brieﬂy list recent work related to the use

of these devices in astronomy. We present in detail

the dataset collected and prepared in Section 3. Fi-

nally, we discuss a potential application with streaks

detection in Section 4, we list additional potential use-

cases in Section 5, and we propose some perspectives

in Section 6.

2 RELATED WORKS

Smartphones capable of producing high-quality astro-

nomical images are fairly recent

, but we can found

references such as (Kim and Ju, 2014) or (Sohib et al.,

2018) in which the authors combine them and optical

devices to enable the practice of EAA. Smartphone

manufacturers have put so much effort into their low-

light photography features that these objects have be-

gun to be of real interest for capturing night sky im-

ages when plugged with other instruments (Falkner,

2020).

Last years, the use of these devices has be-

gun to spread for educational purposes on the sub-

ject of capturing wide-ﬁeld views of constellations

from the Southern Hemisphere (Paula and Micha,

2021), for outreach activities to teach astronomy con-

cepts through Moon phases observation (Malchenko,

2024). In (Fisher and Hinckley, 2022), the authors

show how modern premium smartphones can be com-

pared to Digital Single-Lens Reﬂex (DSLR) to take

astronomical pictures (for instance of Milky Way).

Even NASA has published a detailed guideline to

learn how to use smartphones to observe any kind

of celestial targets

. Nowadays, these devices not

only have sensors but also the on-board computing

capacity to produce, alone or with telescopes support,

formidable results in lunar, solar and deep-sky obser-

vation (L

ecureuil, 2024).

To the best of our knowledge, there is currently no

amateur-accessible database of astronomical images

taken with smartphones, such as a sky survey. In the

next section, we describe a dataset of such images that

we have been collecting and preparing since 2021.

https://en.wikipedia.org/wiki/Camera phone

https://science.nasa.gov/learn/heat/resource/a-guide-t

o-smartphone-astrophotography/

3 DATASET DESCRIPTION

AstroSmartphoneDataset, is a collection of wide-ﬁeld

astronomical images captured with Android smart-

phones by using Pixel Camera (also called Google

Camera or GCam), a Google mobile application al-

lowing to take long-exposure astronomical images

with a large ﬁeld of view

. The astrophotography

mode of this mobile application allows to capture

and produce long-exposure frames to reduce noise,

enhance details, improve dynamic range, while pre-

serving non-astronomical elements like buildings and

landscapes, ensuring a natural look.

In order to test Deep Learning algorithms as part

of research projects carried out at the Luxembourg In-

stitute of Science and Technology (LIST), data have

been collected between 2021 and 2025 from different

locations in France. To do this, we have used four

smartphones, each with different sensors and onboard

calculation capabilities:

• Pixel 4a: 12.2 megapixels rear camera, f/1.7 aper-

ture, HDR+ technology.

• Pixel 6: dual rear cameras, 50 megapixels wide-

angle and 12 megapixels ultra-wide.

• Pixel 8a: dual rear cameras, 64 megapixels wide-

angle and 13 megapixels ultra-wide.

• Pixel 8Pro: triple rear cameras, 50 megapixels

wide-angle, 48 megapixels ultra-wide, and 48

megapixels telephoto.

Data was acquired over a long period of almost four

years, in different regions of France (mainly in the

north-east in a non-urban area, in the north in the

countryside, in the south by the sea, in the centre in

the mountains): these different locations have a Bor-

tle level between 4 and 8 (it is a 1-to-9 rating system

indicating light pollution levels: the lower the num-

ber, the darker the sky and the better the conditions

for stargazing (Bortle, 2001)).

As these smartphones were not released at the

same time and we did not have them at our dis-

posal when we started collecting data, we used them

in chronological order: ﬁrst the Pixel4a, then the

Pixel8Pro, then the Pixel6 and the Pixel8a. The ﬁrst

image was captured on 30 July 2021 with a Pixel4a

(Figure 1) while the last image was taken from 18

March 2025 with a Pixel8a.

Data acquisition required the use of often inge-

nious techniques to orientate and stabilise the phones

during capture. Indeed, the idea was to never have ex-

actly the same point of view, so the smartphone used

was sometimes placed on a chair, on a table, on the

https://en.wikipedia.org/wiki/Pixel Camera

Toward a Public Dataset of Wide-Field Astronomical Images Captured with Smartphones

645

Figure 1: First image of AstroSmartphoneDataset, captured

with a Pixel4a from the North-East of France in a suburban

area (30 July 2021).

ﬂoor, held against another object such as a bottle, etc.

The main constraint was to ensure that the phone was

stable during image capture (requiring the use of the

timer, to avoid shaking after pressing the buttons).

Images were taken under different weather condi-

tions (clear sky, foggy sky, almost overcast sky), in

order to obtain an heterogeneous set. In practice, the

photo mode used means that the stars can be captured

on the uncovered parts of the night sky, even if the

other part is cloudy.

There was no strict curation of collected images:

only really bad images were ignored after capture and

then not included in the dataset (for example, blurry

images due to incorrect use of the smartphones).

As a result, AstroSmartphoneDataset contains

4635 medium-resolution and 197 high-resolution

RGB images in JPEG format (minimal compression),

and each RGB image has a minimal resolution of

1440 × 1920 pixels: they were obtained after reg-

istration, stacking and post processing of one-second

shots, giving ﬁnal images corresponding to a cumula-

tive exposure time of 4 minutes.

The GCam application produces two kinds of

data: high-resolution images and timelapses (under

the form of short MP4 movies) of 1 second. These

timelapses are also the result of post-processing (i.e.

alignment, stacking and HDR), and we use them to

extract additional images, albeit at a lower resolution

than the previous images.

The unique characteristic of the images in the

dataset is attributable to two factors. Firstly, the wide

ﬁeld of view through which the images were captured,

and secondly, the special processing carried out by the

GCam application. The resulting images thus com-

bine elements of astronomical images, characterised

by a ﬁeld of stars, and potentially very bright galaxies

and nebulae, with elements from the landscape at the

time the shot was taken, such as houses, walls, roofs

and trees.

AstroSmartphoneDataset is available as an open

ZIP archive on Zenodo

(Parisot, 2025).

From a purely entertaining point of view, these

wide-ﬁeld images allow you to admire the Milky Way

(our galaxy) at different times of the year, as well

as other parts of the sky, including well-known con-

stellations such as Orion and the famous Big Dipper.

However, we will see in the next few sections that this

data can also be used for other purposes.

4 USE CASE: STREAKS

DETECTION

A notable feature of the AstroSmartphoneDataset

is the signiﬁcant presence of streaks in the high-

resolution images, and particularly in the medium-

resolution images derived from the timelapses

movies. The aforementioned streaks are typically ob-

served in the form of lines exhibiting varying degrees

of thickness and continuity, and are attributed to vari-

ous sources such as satellites, airplanes, and even me-

teors (Nir et al., 2018): these lines are often problem-

atic in astrophotography because they lead to data loss

and generate additional work to remove them (un-

painting).

Traditionally, the detection of such moving ob-

jects is a key challenge in Computer Vision, and

common techniques include: Background Subtrac-

tion (identiﬁes changes between frames by modeling

the background), Optical Flow (Estimates pixel mo-

tion across frames (Farneb

ack, 2003)), Deep Learning

(Convolutional Neural Networks or Recurrent Neu-

ral Networks for feature extraction and tracking). In

(Petit et al., 2004), the authors discuss a method for

detecting moving objects in astronomical images us-

ing photometric variation analysis and image subtrac-

tion techniques: it focuses on improving the detec-

tion of faint sources by optimizing background mod-

https://zenodo.org/records/14933725

DATA 2025 - 14th International Conference on Data Science, Technology and Applications

646

eling and noise reduction. (Lee, 2023) presents a

Deep Learning approach for detecting moving trans-

Neptunian objects (TNOs) in large sky surveys: it

uses Convolutional Neural Networks trained on artiﬁ-

cially generated sources in time-series images from

the CFHT MegaCam. The method presented in

(Huang et al., 2024) is applied to tracking dynamic

space targets, such as satellites and space debris, us-

ing image sequences for motion detection and Deep-

SORT for tracking. In (Parisot and Jaziri, 2024), the

authors combined classiﬁcation with ResNet50 and

GradCAM to detect linear features in raw astronomi-

cal images.

A potential application of the AstroSmartphone-

Dataset is the construction of an annotated dataset,

which can be utilised to train an automatic detec-

tion model based on YOLO (You Only Look Once)

(Jegham et al., 2025). YOLO is a real-time object de-

tection method that analyze an entire image in a sin-

gle pass, identifying and localizing multiple objects

simultaneously with high speed and accuracy.

As a ﬁrst step, we have manually selected images,

we have split them into 640 × 640 patches and we

have annotated the streaks with the MakeSense tool

, and we have stored images and labels by using the

YOLO format: 403 images for training, 88 images for

validation, 163 images for test (Figure 2). Given that

our images contain night sky, landscape features and

streaks to be detected, we made sure to keep back-

ground images (i.e. without labels) in each of these

sets, to show that features that might look like streaks

(house edges, tree branches, etc.) are not.

Then we have trained different versions of YOLO

released between 2022 and 2024 (v7, v8 and v11),

with different sizes (tiny, normal, extra-large), by us-

ing the YOLOv7 ofﬁcial Python package

and the

Ultralytics Python package

without speciﬁc cus-

tomization.

YOLO-v7 implementation uses default hyperpa-

rameters such as lr0 = 0.01, momentum = 0.937,

weight

ecay = 0.0005, with augmentations like

mosaic = 1.0, f liplr = 0.5, and scale = 0.5. YOLO-

v8, from Ultralytics, follows similar defaults but with

slight adjustments and improved augmentation han-

dling. YOLO-v11 extends YOLO-v8 and includes

automatic tuning over hyperparameter ranges to op-

timize training dynamically.

For each type of model, we ran the training in

the same way over several hundred epochs (maximum

500), and in most cases, the best model was found af-

ter a calculation between 150 or 300 epochs.

https://www.makesense.ai

https://github.com/WongKinYiu/yolov7

https://pypi.org/project/ultralytics/

Figure 2: Four 640 × 640 annotated images in the training

set: the last one is for background, i.e. without label, to

learn the model to ignore undesired patterns in the image.

Based on our tests (Table 1), we found that the best

models in terms of accuracy are those with the largest

number of parameters, namely yolo-v11-x (

57M pa-

rameters) and yolo-v7-x (

71M parameters). With

26M parameters, yolo-v8-m completes the podium of

the best models evaluated, according to the test set.

Figure 4 illustrates the predictions of two models

(YOLO-v8-x and YOLO-v11-x) on a sample image

from AstroSmartphoneDataset.

This type of model can be used to directly detect

whether streaks appear during observations, for moni-

toring purposes (Space Domain Awareness) or simply

to ﬁlter/reject impacted images if they should not con-

tain such artefacts. It’s a hot topic: the Centre for the

Protection of the Dark and Quiet Sky from Satellite

Constellation Interference (CPS), established by the

International Astronomical Union (IAU) in 2022, co-

ordinates global efforts to mitigate the impact of satel-

lite constellations on astronomy and the night sky.

5 DISCUSSION

As is, AstroSmartphoneDataset contains only images,

without labels or metadata, and may be useful for

unsupervised or self-supervised learning. It enables

tasks like autoencoding, clustering, contrastive learn-

ing, and Generative Adversarial Networks (GAN). It

can also be leveraged for transfer learning, using pre-

trained models (ResNet) to extract features or gen-

erate pseudo-labels with AI models like Contrastive

Language–Image Pretraining (CLIP) (Radford et al.,

Toward a Public Dataset of Wide-Field Astronomical Images Captured with Smartphones

647

Table 1: Evaluation of the accuracies of the different trained YOLO models – computation was done on the test set (163

images). For each model, the count of parameters is also speciﬁed.

Model Parameters Precision Recall mAP@.5 mAP@.95

yolo-v7 37M 0.553 0.514 0.444 0.187

yolo-v7-tiny 6M 0.575 0.495 0.415 0.146

yolo-v7-x 71M 0.588 0.654 0.568 0.247

yolo-v8-n 3M 0.533 0.495 0.43 0.176

yolo-v8-m 26M 0.474 0.488 0.491 0.224

yolo-v8-x 68M 0.422 0.514 0.412 0.159

yolo-v11-n 2.5M 0.448 0.637 0.492 0.229

yolo-v11-m 20M 0.399 0.486 0.364 0.151

yolo-v11-x 57M 0.704 0.626 0.646 0.332

Figure 3: Statistics obtained after the training during 400 epochs of the YOLO-v8-m model with the default hyper-parameters

of the Ultralytics Python package.

2021).

Other use cases can be carried out on AstroSmart-

phoneDataset, with an additional annotation effort (as

for the detection of streaks presented in the previous

section):

• One example is the possibility of creating an im-

age segmentation model, to be able to clearly dis-

tinguish what is sky and what is landscape: YOLO

can also be used for this type of task (Kang and

Kim, 2023).

• Another example is the design of an Image Qual-

ity Assessment (IQA) system for this kind of

astronomical images captured with smartphones

(IQA), again taking into account the hybrid as-

pect of images: we could imagine a No Refer-

ence IQA system that assesses both the quality of

the sky (shape of the stars, darkness of the back-

ground, etc.) and the quality of the foreground or

background elements: a technique like CLIP-IQA

could be a good basis (Wang et al., 2023).

Another potential application of AstroSmartphone-

Dataset is research about light pollution. By collect-

ing images from locations all over the world, users

can contribute to a global analysis of how artiﬁ-

cial light impacts visibility in the night sky (Mu

noz-

Gil et al., 2022). This would assist citizen science

projects and environmental studies of light pollution.

The large number of images captured, can potentially

enable scientists and environmental groups to better

understand the patterns and extent of light pollution,

a task that would be difﬁcult with stationary and tra-

ditional equipment. This would allow amateur as-

tronomers and citizen scientists to contribute together

to such global initiatives.

All these perspectives open the door to the cre-

ation of new processing methods for the data pro-

DATA 2025 - 14th International Conference on Data Science, Technology and Applications

648

Figure 4: Example of inferences on 640 × 640 images with

the trained yolo-8-x and yolo-11-x models: the second one

is more efﬁcient to ﬁnd correctly the bounding boxes around

the streaks.

duced by a mechanism similar to that used to generate

the images in this dataset.

There is one particular limitation, speciﬁc to

datasets for Machine Learning/Deep Learning pur-

poses: the models trained/validated on AstroSmart-

phoneDataset will only be valid and effective if they

are applied to similar images, i.e. captured with the

same types of smartphone, or at least with similar

characteristics (sensors, etc.), but also with similar

shooting conditions.

6 CONCLUSION AND

PERSPECTIVES

In this paper, we have introduced and described As-

troSmartphoneDataset, a collection of wide-ﬁeld as-

tronomical images captured with various smartphones

since 2021 from different locations in France. This

dataset represents a valuable resource for studying

and improving image processing techniques for as-

tronomical photography on mobile devices. By pro-

viding a diverse and representative dataset of the cur-

rent capabilities of these devices, we pave the way for

new advancements in Deep Learning applied to mo-

bile astrophotography. We hope AstroSmartphone-

Dataset will encourage the scientiﬁc community to

explore innovative approaches for other Computer Vi-

sion tasks, object recognition, auto-labelling in im-

ages captured with smartphones. Thus, we have pro-

posed a use-case consisting in labelling the images to

train YOLO models for streaks detection, enabling to

automatically track transient streak-like objects such

as satellites or meteors across large datasets.

In future work, we will continue to capture ad-

ditional images with other types of smartphones and

from other locations, to to improve dataset generaliz-

ability. Moreover, we will use the data to design and

train tiny Deep Learning models, with a view to em-

bed them directly in smartphones: this could enable

real-time inference, allowing users to instantly ana-

lyze their astrophotography images without the need

for powerful computing infrastructures.

ACKNOWLEDGEMENTS

This research was funded by the Luxembourg Insti-

tute of Science and Technology (LIST). Data storage,

data processing and Deep Learning models training

were conducted on the LIST Artiﬁcial Intelligence

and Data Analytics platform (AIDA), with the assis-

tance of Jean-Franc¸ois Merche and Raynald Jadoul.

DATA AVAILABILITY

AstroSmartphoneDataset is available as a ZIP archive

on Zenodo (https://zenodo.org/records/14933725),

and a dedicated Github repository has also been set to

aggregate related content (https://github.com/opari

sot/AstroSmartphoneDataset). Additional materials

used to support the results of this paper are available

from the corresponding author upon request.

Toward a Public Dataset of Wide-Field Astronomical Images Captured with Smartphones

649

REFERENCES

Arimatsu, K. (2025). The OASES project: exploring the

outer Solar System through stellar occultation with

amateur-class telescopes. Philosophical Transactions

A, 383(2291):20240191.

Ashley, J. (2017). Video Astronomy on the Go. Springer

International Publishing.

Bortle, J. E. (2001). The bortle dark-sky scale. Sky and

Telescope, 161(126):5–6.

Drechsler, M., Strottner, X., Sainty, Y., Fesen, R. A.,

Kimeswenger, S., Shull, J. M., Falls, B., Vergnes, C.,

Martino, N., and Walker, S. (2023). Discovery of ex-

tensive [o iii] emission near m31. Research Notes of

the AAS, 7(1):1.

Falkner, D. E. (2020). Astrophotography using a compact

digital camera or smartphone camera. In The Mythol-

ogy of the Night Sky, pages 269–294. Springer.

Fang, Z., Ignatov, A., Zamﬁr, E., and Timofte, R. (2023).

Sqad: Automatic smartphone camera quality assess-

ment and benchmarking. In Proceedings of the

IEEE/CVF International Conference on Computer Vi-

sion, pages 20532–20542.

Farneb

ack, G. (2003). Two-frame motion estimation based

on polynomial expansion. In Image Analysis: 13th

Scandinavian Conference, SCIA 2003 Halmstad, Swe-

den, June 29–July 2, 2003 Proceedings 13, pages 363–

370. Springer.

Fisher, A. and Hinckley, S. (2022). Smartphone astropho-

tography. In Proceedings of the IUPAP International

Conference on Physics Education 2022, pages 84–84.

Huang, C., Zeng, Q., Xiong, F., and Xu, J. (2024).

Space dynamic target tracking method based on ﬁve-

frame difference and deepsort. Scientiﬁc Reports,

14(1):6020.

Jegham, N., Koh, C. Y., Abdelatti, M., and Hendawi,

A. (2025). YOLO Evolution: A Comprehensive

Benchmark and Architectural Review of YOLOv12,

YOLO11, and Their Previous Versions.

Kang, C. H. and Kim, S. Y. (2023). Real-time object detec-

tion and segmentation technology: an analysis of the

yolo algorithm. JMST Advances, 5(2):69–76.

Kim, J.-H. and Ju, Y.-G. (2014). Fabrication of a video tele-

scope using a smartphone. New Physics, 64(3):307–

312.

ecureuil, P. (2024). Photographier l’Univers avec

un smartphone: Petit guide d’initiation

l’astrophotographie. De Boeck Sup

erieur.

Lee, A. (2023). A Journey to the Edge of the Solar Sys-

tem with an AI navigator. PhD thesis, University of

Victoria.

Malchenko, S. L. (2024). From smartphones to stargazing:

the impact of mobile-enhanced learning on astronomy

education. Science Education Quarterly, 1(1):1–7.

noz-Gil, G., Dauphin, A., Beduini, F. A., and S

anchez de

Miguel, A. (2022). Citizen science to assess light

pollution with mobile phones. Remote Sensing,

14(19):4976.

Nir, G., Zackay, B., and Ofek, E. O. (2018). Optimal and

efﬁcient streak detection in astronomical images. The

Astronomical Journal, 156(5):229.

Parisot, O. (2025). AstroSmartphoneDataset: a collection

of astronomical images captured with smartphones.

Parisot, O., Bruneau, P., Hitzelberger, P., Krebs, G., and

Destruel, C. (2022). Improving accessibility for deep

sky observation. ERCIM News, 2022(130).

Parisot, O. and Jaziri, M. (2024). Impact of Satellites

Streaks for Observational Astronomy: A Study on

Data Captured During One Year from Luxembourg

Greater Region. In Proceedings of the 13th Interna-

tional Conference on Data Science, Technology and

Applications - Volume 1: DATA, pages 417–424. IN-

STICC, SciTePress.

Paula, M. E. and Micha, D. N. (2021). Low-cost astropho-

tography with a smartphone: Steam in action. The

Physics Teacher, 59(6):446–449.

Petit, J.-M., Holman, M., Scholl, H., Kavelaars, J., and

Gladman, B. (2004). A highly automated moving ob-

ject detection package. Monthly Notices of the Royal

Astronomical Society, 347(2):471–480.

Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G.,

Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark,

J., et al. (2021). Learning transferable visual models

from natural language supervision. In International

conference on machine learning, pages 8748–8763.

PmLR.

Romanov, F. (2022). The contribution of the modern am-

ateur astronomer to the science of astronomy. arXiv

preprint arXiv:2212.12543.

Sohib, A., Danusaid, N., Sawitri, A., Nuryadin, B. W.,

and Agustina, R. D. (2018). Low cost digitaliza-

tion of observation telescope by utilizing smartphone.

In MATEC Web of Conferences, volume 197, page

02002. EDP Sciences.

Wang, J., Chan, K. C., and Loy, C. C. (2023). Exploring clip

for assessing the look and feel of images. In Proceed-

ings of the AAAI conference on artiﬁcial intelligence,

volume 37, pages 2555–2563.

Woodhouse, C. (2015). The Astrophotography Manual.

Routledge.

DATA 2025 - 14th International Conference on Data Science, Technology and Applications

650