Astronomical image time series classification using CONVolutional attENTION (ConvEntion)

Aims. The treatment of astronomical image time series has won increasing attention in recent years. Indeed, numerous surveys following up on transient objects are progress or under construction, such as the Vera C. Rubin Observatory Legacy Survey for Space and Time (LSST), which is poised to produce huge amounts these series. associated scientific topics extensive, ranging from study our galaxy observation most distant supernovae measuring expansion universe. With a large amount data available, need robust automatic tools detect classify celestial growing steadily. Methods. This based assumption that images contain more information than light curves. In this paper, we propose novel approach deep learning classifying different types space directly using images. We named ConvEntion, stands CONVolutional attENTION. It convolutions transformers, new approaches Our solution integrates spatio-temporal features can be applied various datasets with any number bands. Results. work, solved problems tend suffer present results classifications an increase accuracy 13%, compared state-of-the-art use series, 12% increase,