Finding prominent objects in images without manual labeled training
Finding prominent objects in images without manual labeled training
Large-Scale Unsupervised Object Discovery
arXiv paper abstract https://arxiv.org/abs/2106.06650v1
arXiv PDF paper https://arxiv.org/pdf/2106.06650v1.pdf
GitHub https://github.com/huyvvo/LOD (paper says code will be released there but nothing there on 6/22/21)
Existing approaches to unsupervised object discovery (UOD) do not scale up to large datasets without approximations which compromise their performance.
… propose … formulation of UOD as a ranking problem, amenable to … methods … for eigenvalue problems and link analysis.
… where a single prominent object is sought … approach is … over 37% better than … other algorithms capable of … 1.7M images.
… where multiple objects are sought … over 14% better in average precision …
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website
LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b
