AI智能圖片擷取爬蟲開發者（針對電商產品頁面，辨識與過濾非目標圖）

需求說明

We are looking for a high-end AI image analysis and web crawler developer to be responsible for developing a "product image capture system" that can intelligently judge image content. This system will be integrated into our AI try-on platform (Dukira). The goal is to automatically intercept clean and model-free product pictures from major e-commerce websites, and automatically exclude non-target pictures. This is not a general crawler task. You need to know how to handle image classification, AI screening logic and HTML structure analysis. ⸻ Key tasks of work: • Automatically capture product page pictures of e-commerce websites (such as clothes, glasses, bags, etc.) • Automatically identify and filter the photos of models • Exclude the pictures of recommended areas, footers or other non-main products • Keep "clean main product pictures" (no people, no distractions) • After the picture is captured, it can be automatically sent to our next scanning processing system (API transmission) ⸻ Skill requirements: • Familiar with Python and Scrapy, Playwright, or Puppeteer • Experience in AI image classification (can be used for ResNet, YOLO, CLIP, or self-training model) • It can integrate computer visual models to judge whether the picture contains characters or other products. • Familiar with the structure of e-commerce web pages, and can find out the main map block from HTML/CSS • It can be encapsulated into modules and automatically batch process multiple websites. • Support API output or save to database (such as MongoDB) ⸻ Conditions for adding points: • Have experience in dealing with crawlers on clothing e-commerce type pages • Familiar with Cloudflare / Anti-crawling technology bypass • The whole process can be deployed to AWS or Lambda ⸻ Budget and cooperation mode: • Budget: NT$15,000 • If the results are excellent, there will be long-term cooperation opportunities in the future. • Need to provide testable Demo or GitHub examples ⸻ How to strip: Please provide the following information: 1. Introduction of relevant experience (especially the part of picture identification and e-commerce crawlers) 2. Links of past projects or works (GitHub, videos, Demo, etc.)