{"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/an-annotation-saved-is-an-annotation-earned","title":"An Annotation Saved is an Annotation Earned: Using Fully Synthetic Training for Object Instance Detection","arxiv_id":"1902.09967","date":"2019-02-26","proceeding":null,"authors":["Stefan Hinterstoisser","Olivier Pauly","Hauke Heibel","Martina Marek","Martin Bokeloh"],"abstract":"Deep learning methods typically require vast amounts of training data to\nreach their full potential. While some publicly available datasets exists,\ndomain specific data always needs to be collected and manually labeled, an\nexpensive, time consuming and error prone process. Training with synthetic data\nis therefore very lucrative, as dataset creation and labeling comes for free.\nWe propose a novel method for creating purely synthetic training data for\nobject detection. We leverage a large dataset of 3D background models and\ndensely render them using full domain randomization. This yields background\nimages with realistic shapes and texture on top of which we render the objects\nof interest. During training, the data generation process follows a curriculum\nstrategy guaranteeing that all foreground models are presented to the network\nequally under all possible poses and conditions with increasing complexity. As\na result, we entirely control the underlying statistics and we create optimal\ntraining samples at every stage of training. Using a set of 64 retail objects,\nwe demonstrate that our simple approach enables the training of detectors that\noutperform models trained with real data on a challenging evaluation dataset.","url_abs":"http://arxiv.org/abs/1902.09967v1","url_pdf":"http://arxiv.org/pdf/1902.09967v1.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":"an-annotation-saved-is-an-annotation-earned","repo_url":"https://github.com/Unity-Technologies/SynthDet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1902.09967","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}