I’m building a deep‑learning model that classifies images as authentic or forged, and I need a corpus of real “fake” images pulled directly from social‑media platforms. So far I’ve: Searched existing datasets (e.g. CASIA) – they’re useful but not sourced from social media. Tried keyword scraping via the Twitter API (searching for terms like “Photoshop” or “deepfake”) – results are sparse and low quality. Looked at image‑forensics repositories (GitHub, Kaggle) – again, none are specifically harvested from Instagram, TikTok, Facebook, etc. My goal is to automate collection of user‑generated forged images (Photoshop edits, deepfakes, memes, etc.) directly from social‑media streams. What I’m asking: Are there recommended APIs, tools or scraping frameworks specialized for finding/managing “fake” images on social platforms? Best practices for filtering and verifying that an image is truly forged before adding it to my dataset? Any existing public sources or communities that share user‑created forgeries suitable for ML training? I’m open to Python‑based solutions, usage quotas/limits, and any pointers on legal/ethical scraping. Continue reading...
