The weights of the models trained on the task of quiremark detection on medieval manuscripts. Specifically this archive contains the model weights of two detection approaches used in our work represented by different architectures (Faster RCNN and ViT), each of which is further represented by a set of five weights corresponding to the resulting models trained when performing the final evaluation on all 59 manuscripts in the form of 5-fold cross validation. The specific manuscripts used to train/eval the models in each fold are also specified in this archive.