We propose an automatic pipeline for organizing, ranking, and suggesting the best images in a user’s
gallery based on computer vision methods on constrained devices. In contrast to most present solutions,
our current implementation clusters the images using a hierarchical approach based on three levels of
information (GPS location, datetime and image content) and suggests the best images based on
aesthetical and technical scores using a hybrid network. This compact network shares the same backbone
for the scores prediction while also being used to extract features from the images, which are then used
to clustering. This makes the solution lightweight, favoring constrained devices, while also enabling
running the entire pipeline locally, ensuring the user’s privacy.
Creative Commons License
This work is licensed under a Creative Commons Attribution-Share Alike 4.0 License.
INC, HP, "HIERARCHICAL MULTIMODAL CLUSTERING USING A MULTI-TASK NETWORK WITH HIGHLIGHT SELECTION: A NOVEL APPROACH TO ORGANIZE AN IMAGE GALLERY", Technical Disclosure Commons, (March 01, 2021)