{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/multi-stage-variational-auto-encoders-for","title":"Multi-Stage Variational Auto-Encoders for Coarse-to-Fine Image Generation","arxiv_id":"1705.07202","date":"2017-05-19","proceeding":null,"authors":["Lei Cai","Hongyang Gao","Shuiwang Ji"],"abstract":"Variational auto-encoder (VAE) is a powerful unsupervised learning framework\nfor image generation. One drawback of VAE is that it generates blurry images\ndue to its Gaussianity assumption and thus L2 loss. To allow the generation of\nhigh quality images by VAE, we increase the capacity of decoder network by\nemploying residual blocks and skip connections, which also enable efficient\noptimization. To overcome the limitation of L2 loss, we propose to generate\nimages in a multi-stage manner from coarse to fine. In the simplest case, the\nproposed multi-stage VAE divides the decoder into two components in which the\nsecond component generates refined images based on the course images generated\nby the first component. Since the second component is independent of the VAE\nmodel, it can employ other loss functions beyond the L2 loss and different\nmodel architectures. The proposed framework can be easily generalized to\ncontain more than two components. Experiment results on the MNIST and CelebA\ndatasets demonstrate that the proposed multi-stage VAE can generate sharper\nimages as compared to those from the original VAE.","url_abs":"http://arxiv.org/abs/1705.07202v1","url_pdf":"http://arxiv.org/pdf/1705.07202v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"multi-stage-variational-auto-encoders-for","repo_url":"https://github.com/divelab/msvae","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":"image-generation","task_name":"Image Generation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1705.07202","atlas_url":"https://app.syntology.ai/?focus=1705.07202","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}