About 30\% of disk galaxies show lopsidedness in their stellar disk. Although such a large-scale asymmetry in the disk can be primarily looked upon as a long-lived mode (m=1), the physical origin of the lopsidedness in the disk continues to be a puzzle. In this work, we develop an automated approach to identify lopsided galaxies from the SDSS DR18 using a Deep Convolutional Neural Network (DCNN) based on the publicly available AlexNet architecture. We select nearly face-on spiral galaxies from SDSS DR18 with the Petrosian 90\% light radius (petroR90\ᵢ) greater than 20^''. Based on the visual inspection, we choose 106 lopsided spiral galaxies and 105 symmetric spiral galaxies, as our training set. Our trained model achieves a testing accuracy of 92. 8\% at the end of 150 epochs. We then employ the trained model on a set of 813 face-on spiral galaxies from SDSS DR18 with 17^'' petroR90\ᵢ 20^'' and identify 452 new lopsided spiral galaxies. We next investigate the cosmic web environments in which the galaxies are located, using the Hessian matrix of the density field. We find that 39\% of the lopsided galaxies are located in sparser environments such as sheets and voids. This may provide interesting clues towards understanding the origin of lopsidedness in isolated galaxies, where distortion due to the tidal interactions is less frequent.
Saha et al. (Mon,) studied this question.