{"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/learning-to-compress-and-search-visual-data","title":"Learning to compress and search visual data in large-scale systems","arxiv_id":"1901.08437","date":"2019-01-24","proceeding":null,"authors":["Sohrab Ferdowsi"],"abstract":"The problem of high-dimensional and large-scale representation of visual data\nis addressed from an unsupervised learning perspective. The emphasis is put on\ndiscrete representations, where the description length can be measured in bits\nand hence the model capacity can be controlled. The algorithmic infrastructure\nis developed based on the synthesis and analysis prior models whose\nrate-distortion properties, as well as capacity vs. sample complexity\ntrade-offs are carefully optimized. These models are then extended to\nmulti-layers, namely the RRQ and the ML-STC frameworks, where the latter is\nfurther evolved as a powerful deep neural network architecture with fast and\nsample-efficient training and discrete representations. For the developed\nalgorithms, three important applications are developed. First, the problem of\nlarge-scale similarity search in retrieval systems is addressed, where a\ndouble-stage solution is proposed leading to faster query times and shorter\ndatabase storage. Second, the problem of learned image compression is targeted,\nwhere the proposed models can capture more redundancies from the training\nimages than the conventional compression codecs. Finally, the proposed\nalgorithms are used to solve ill-posed inverse problems. In particular, the\nproblems of image denoising and compressive sensing are addressed with\npromising results.","url_abs":"http://arxiv.org/abs/1901.08437v1","url_pdf":"http://arxiv.org/pdf/1901.08437v1.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":"learning-to-compress-and-search-visual-data","repo_url":"https://github.com/sssohrab/PhDthesis","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"learning-to-compress-and-search-visual-data","repo_url":"https://github.com/CorentinBT/Decoder","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"learning-to-compress-and-search-visual-data","repo_url":"https://github.com/sssohrab/sparsifying_groups_imAmbiguation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"compressive-sensing","task_name":"Compressive Sensing"},{"task_slug":"denoising","task_name":"Denoising"},{"task_slug":"image-compression","task_name":"Image Compression"},{"task_slug":"image-denoising","task_name":"Image Denoising"},{"task_slug":"retrieval","task_name":"Retrieval"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}