The Pareidolia Paradox

The Pareidolia Paradox is an online machine learning challenge focused on image classification and computer vision. Participants must classify 256x256 grayscale lunar surface images into one of two categories: Class 0 (Depth: Craters, holes, and surface depressions) or Class 1 (Rise: Mounds, hills, rocks, and boulders).

The challenge addresses the phenomenon of topographic inversion, where lunar terrain appearance changes with sunlight direction. Each image includes a sun_azimuth_angle in its metadata, which participants are expected to use. The dataset consists of 7,854 training images and 2,000 evaluation images, with metadata provided in CSV files. Submissions require a single CSV file with 'image_id,label' and will be evaluated using Balanced Accuracy.