{"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/treepedia-20-applying-deep-learning-for-large","title":"Treepedia 2.0: Applying Deep Learning for Large-scale Quantification of Urban Tree Cover","arxiv_id":"1808.04754","date":"2018-08-14","proceeding":null,"authors":["Bill Yang Cai","Xiaojiang Li","Ian Seiferling","Carlo Ratti"],"abstract":"Recent advances in deep learning have made it possible to quantify urban\nmetrics at fine resolution, and over large extents using street-level images.\nHere, we focus on measuring urban tree cover using Google Street View (GSV)\nimages. First, we provide a small-scale labelled validation dataset and propose\nstandard metrics to compare the performance of automated estimations of street\ntree cover using GSV. We apply state-of-the-art deep learning models, and\ncompare their performance to a previously established benchmark of an\nunsupervised method. Our training procedure for deep learning models is novel;\nwe utilize the abundance of openly available and similarly labelled\nstreet-level image datasets to pre-train our model. We then perform additional\ntraining on a small training dataset consisting of GSV images. We find that\ndeep learning models significantly outperform the unsupervised benchmark\nmethod. Our semantic segmentation model increased mean intersection-over-union\n(IoU) from 44.10% to 60.42% relative to the unsupervised method and our\nend-to-end model decreased Mean Absolute Error from 10.04% to 4.67%. We also\nemploy a recently developed method called gradient-weighted class activation\nmap (Grad-CAM) to interpret the features learned by the end-to-end model. This\ntechnique confirms that the end-to-end model has accurately learned to identify\ntree cover area as key features for predicting percentage tree cover. Our paper\nprovides an example of applying advanced deep learning techniques on a\nlarge-scale, geo-tagged and image-based dataset to efficiently estimate\nimportant urban metrics. The results demonstrate that deep learning models are\nhighly accurate, can be interpretable, and can also be efficient in terms of\ndata-labelling effort and computational resources.","url_abs":"http://arxiv.org/abs/1808.04754v1","url_pdf":"http://arxiv.org/pdf/1808.04754v1.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":"treepedia-20-applying-deep-learning-for-large","repo_url":"https://github.com/billcai/treepedia_dl_public","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"deep-learning","task_name":"Deep Learning"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}