18. Dezember 2025
For ‘The Monk by the Sea’ by Caspar David Friedrich, you will initially find 18 duplicates in the image archive under ‘Similar images’Your text to link here… because they have the smallest vector distance in the image similarity search.
What is behind this? For the image similarity search, a method using image vectors created with the help of the self-supervised learning algorithm SwAV (Swapping Assignments between Views) was integrated. The implementation comes from Facebook, using a pre-trained model on the ImageNet dataset (ImageNet). The resulting vectors were reduced to 80 dimensions, calculated for all images in the archive and stored in the index. Search queries are based on comparing these vectors using Euclidean distance – this displays particularly similar images.
