AutoCrawler

Why we included this project

Assembling a decent image set for model training usually means saving search results one at a time, and AutoCrawler automates that chore. It drives a real Chrome browser through Selenium to collect matching images from Google and Naver, spreading the work across multiple threads so a long keyword list doesn't take all day. You can ask for full-resolution files instead of thumbnails, enable face-focused searches, and it flags any folder that comes in well under the average image count, which is a quick way to catch an under-filled class before training starts. The command line exposes thread count, headless mode, proxy lists, and per-site download caps, and when Google changes its page layout you patch collect_links.py rather than the whole tool. For anyone building custom vision datasets who wants a scriptable, repeatable way to pull thousands of reference images, this gets the job done.