{"id":17572369,"url":"https://github.com/eyewheel/imgrep","last_synced_at":"2026-02-09T13:32:07.166Z","repository":{"id":258860825,"uuid":"875316969","full_name":"eyewheel/imgrep","owner":"eyewheel","description":"full text search your meme/screenshot folder with small LMs x traditional OCR","archived":false,"fork":false,"pushed_at":"2024-11-19T07:52:14.000Z","size":28,"stargazers_count":3,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"master","last_synced_at":"2025-06-14T19:09:09.439Z","etag":null,"topics":["moondream","ocr","ollama","ollama-api","small-language-models"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/eyewheel.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE.md","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-10-19T16:52:23.000Z","updated_at":"2025-03-22T17:20:23.000Z","dependencies_parsed_at":"2025-03-06T09:46:19.700Z","dependency_job_id":null,"html_url":"https://github.com/eyewheel/imgrep","commit_stats":{"total_commits":17,"total_committers":1,"mean_commits":17.0,"dds":0.0,"last_synced_commit":"08bf479fe4d868fd70db52dc055c93cb6881e06c"},"previous_names":["eyewheel/imgrep"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/eyewheel/imgrep","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/eyewheel%2Fimgrep","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/eyewheel%2Fimgrep/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/eyewheel%2Fimgrep/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/eyewheel%2Fimgrep/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/eyewheel","download_url":"https://codeload.github.com/eyewheel/imgrep/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/eyewheel%2Fimgrep/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":29266947,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-02-09T12:53:16.161Z","status":"ssl_error","status_checked_at":"2026-02-09T12:52:30.244Z","response_time":56,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"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":["moondream","ocr","ollama","ollama-api","small-language-models"],"created_at":"2024-10-21T19:04:02.005Z","updated_at":"2026-02-09T13:32:07.152Z","avatar_url":"https://github.com/eyewheel.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# imgrep\n\n### what does it do?\n\nTranscribes a folder of images containing non-standard text - your folder of memes, screenshots, etc., and makes them searchable.\n\n### how does it work?\n\nGrabs one transcription with a small multimodal model (Moondream, 2B parameters) and one transcription with conventional OCR (pytesseract), and uses the average of similarity to both to search.\n\nThe model can reconstruct deep-fried lines of text or text made difficult to OCR by being superimposed over images; the OCR handles the majority of cases. Most images containing text can be transcribed decently enough to appear in search by at least one of them. Not a pretty solution, but has a pretty high success rate.\n\n### how well does it work?\n\nIn active development. Basic functionality proven to work. Bugs abound.\n\n### how fast does it work?\n\nbenchmarks coming Soon(TM)\n\n### how do I install it?\n\n```\ncurl -fsSL https://ollama.com/install.sh | sh\npip install -r requirements.txt\n```\n\n### how do I run it?\n\n```\nollama run moondream\npython watcher_in_darkness.py\n```\n\nand then move or save your images into WATCH_PATH.\n\n```\npython search.py Text to query for\n```\n\nyes, you can run both simultaneously.\n\n### how much compute do I need?\n\nLess than you think. I've run Moondream inference CPU-only on a $80 laptop and taken ~30 seconds per image. Not optimal, wouldn't recommend, but it works.\n\n### aren't there better open source models?\n\nmuch better! for instance, llama3.2-vision consistently extracts text and responds with only text, even on screenshots of long paragraphs. But llama3.2-vision also takes up 4x as much space and significantly more compute and time.\n\nthis piece of software was assembled specifically to get a working and resilient pipeline without being able to rely on having the best currently available model which can accurately follow criteria - if you have the GPUs on hand to run llama3.2-vision on everything you want to index, you already have access to much better solutions.\n\nthis isn't a trivial fix either that can be improved by using a *slightly* larger  model - llava-7b, for instance, performs worse than moondream across a wide variety of images when asked to extract text only.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Feyewheel%2Fimgrep","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Feyewheel%2Fimgrep","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Feyewheel%2Fimgrep/lists"}