This is a synthetic dataset composed of images containing various ArUco markers overlaid on backgrounds sampled from the MS COCO 2017 training dataset, used to train the models in DeepArUco++: improved detection of square fiducial markers in challenging lighting conditions. The contents of each file are the following: flyingarucov2.tar.gz: Base FlyingArUco v2 images, without any augmentation applied. For each image a .json file is provided with the ground-truth positions of each marker and encoded IDs. det_luma.tar.gz: Detection dataset built from the base FlyingArUco v2 dataset, using only (simulated) changes in lighting in order to obtain augmented samples. det_luma_b.tar.gz: Detection dataset built from the base FlyingArUco v2 dataset, using (simulated) changes in lighting and blur in order to obtain augmented samples. det_luma_bc.tar.gz: Detection dataset built from the base FlyingArUco v2 dataset, using (simulated) changes in lighting, blur and color shift in order to obtain augmented samples. det_luma_bn.tar.gz: Detection dataset built from the base FlyingArUco v2 dataset, using (simulated) changes in lighting, blur and gaussian noise in order to obtain augmented samples. det_luma_bnc.tar.gz: Detection dataset built from the base FlyingArUco v2 dataset, using (simulated) changes in lighting, blur, gaussian noise and color shift in order to obtain augmented samples. reg_luma_bc.tar.gz: Corner refinement (regression)/marker decoding dataset built from the det_luma_bc detection dataset.