SOHO LASCO C2/C3 dataset for training CNN network to classify CME images

This dataset is downloaded from the SOHO LASCO CME CATALOG (SOHO LASCO CME CATALOG - CDAW DATA CENTER (nasa.gov)).We manage to acquire the CME event records (including when the CME appears, ends and its remark) from the CDAW CME catalog websites and download every day running-difference images of LASCO C2 and C3 coronagraph from 2013 to 2018. According to the record of each CME event, an iteration is performed over each image.

If the exposure time of an image is between the appearing time and ending time of an entry from the records, the image is labeled with the corresponding remark of that entry. All images are carefully checked and images with distortions or abnormalities will be removed from dataset to ascertain the quality of the images. The images are 512*512 image files with pixel values ranging in [0,255].