Catalog of High-Confidence Warped and Non-Warped Edge-on Galaxies from DESI Legacy Imaging Surveys DR8

This dataset contains catalogs of high-confidence warped and non-warped edge-on galaxies identified from the DESI Legacy Imaging Surveys using the deep learning classification model developed in this work. Catalogs generated with two confidence thresholds (0.99p≥0.99 and p≥0.9) are provided for both warped and non-warped galaxies. The repository also includes the best-performing trained model weights (.pth) and the labeled galaxy images used for training\validating\testing after the 3D binning procedure. If you use this dataset in your research, please cite both this dataset and the associated journal article (https://arxiv.org/abs/2511.09518). Two catalogs are provided: high_confidence_warp_catalog_pge0p99.csv high_confidence_nonwarp_catalog_pge0p99.csv confidence_warp_catalog_pge0p9.csv confidence_nonwarp_catalog_pge0p9.csv Only galaxies with a maximum predicted classification probability of P ≥ 0.99 or P ≥ 0.9 are included. Each catalog contains the following columns: filename – image filename and unique object identifier; predicted_label – predicted galaxy class (0: non-warped, 1: warped); prob_0 – predicted probability of the non-warped class; prob_1 – predicted probability of the warped class; max_prob – maximum predicted probability; entropy – prediction entropy, representing the uncertainty of the classification. This dataset accompanies the manuscript describing the construction of the catalog and the deep learning classification method. For the labeled galaxy images: The image filenames are formatted as index_DR8ID_RA_Dec.jpg. For example, the filename 12345_396278_123_150.123456_2.345678.jpg corresponds to the DR8 identifier 396278_123. So dr8_id of the galaxy in the Galaxy Zoo DESI catalog can be recovered directly from the image filename.