{"id":22564074,"url":"https://github.com/markmbaum/hirise-thumbnails","last_synced_at":"2025-09-21T17:40:31.996Z","repository":{"id":48412337,"uuid":"503781148","full_name":"markmbaum/hirise-thumbnails","owner":"markmbaum","description":"scripts for scraping a large batch of mars surface imagery from the HiRISE camera","archived":false,"fork":false,"pushed_at":"2022-09-30T01:50:21.000Z","size":10924,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-07-22T09:50:00.087Z","etag":null,"topics":["convolutional-neural-networks","image-processing","mars"],"latest_commit_sha":null,"homepage":"","language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/markmbaum.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null}},"created_at":"2022-06-15T13:38:36.000Z","updated_at":"2022-12-20T06:46:14.000Z","dependencies_parsed_at":"2023-01-18T20:31:23.103Z","dependency_job_id":null,"html_url":"https://github.com/markmbaum/hirise-thumbnails","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/markmbaum/hirise-thumbnails","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/markmbaum%2Fhirise-thumbnails","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/markmbaum%2Fhirise-thumbnails/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/markmbaum%2Fhirise-thumbnails/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/markmbaum%2Fhirise-thumbnails/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/markmbaum","download_url":"https://codeload.github.com/markmbaum/hirise-thumbnails/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/markmbaum%2Fhirise-thumbnails/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":276277582,"owners_count":25615087,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","status":"online","status_checked_at":"2025-09-21T02:00:07.055Z","response_time":72,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["convolutional-neural-networks","image-processing","mars"],"created_at":"2024-12-07T23:14:01.230Z","updated_at":"2025-09-21T17:40:31.977Z","avatar_url":"https://github.com/markmbaum.png","language":"Jupyter Notebook","funding_links":[],"categories":[],"sub_categories":[],"readme":"# HiRISE Thumbnails\n\nThis repo contains a few short script for scraping orbital images of the surface of Mars and associated metadata. Specifically, the images are scraped from the [University of Arizona Lunar \u0026 Planetary Laboratory's catalog](https://www.uahirise.org/catalog/) of interesting/notable targets. Original images are very large, so only the reduced resolution thumbnails are scraped, which are only a couple hundred pixels in each dimension and usually have 3 color channels.\n\nTwo batches of images can be downloaded:\n\n- the `captioned` batch includes about 2,000 images with human-written captions\n- the `non-captioned` batch includes about 70,000 images without human-written captions.\n\nThe data and further description are available [on kaggle](https://www.kaggle.com/datasets/markmbaum/mars-surface-images). There is also a [short notebook](hirise_thumbnails_prep.ipynb) demonstrating how to load the data+metadata, attmepting to mine the image titles for classification labels, and saving cleaned image data to file.\n\nFinally, there is [another notebook](hirise_thumbnails_modeling.ipynb) where I train a convolutional neural network to predict metadata fields from the images using a convolutional neural network. Specifically, I see whether the solar azimuth angle, which is the angle of the sun over the horizon, can be accurately predicted for these images. Possibly, I thought, the model could learn geometric relationships between topographic features and their shadows. In principle this seems feasible and should be easier for Martian images than Earth images. Earth terrain is complicated by vegetation, cloud shadows, human-made structures, and other stuff.\n\nThe model mostly fails, but it **does manage to capture about 20 % of the solar azimuth's variance**. I'm very critical of this performance in the [notebook](hirise_thumbnails_modeling.ipynb), but thinking about it later, I'm mildly impressed that this relatively simple and shallow network grokked the regression at all, even if poorly.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmarkmbaum%2Fhirise-thumbnails","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fmarkmbaum%2Fhirise-thumbnails","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmarkmbaum%2Fhirise-thumbnails/lists"}