Hey, I am trying to pass some images to gpt-4-vision but I am unable to parse some images. Especially the ones from the website “publicdomainpictures”
For other applications, to scrape the image, it is possible by modifying the headers and user agents. Is such a thing possible in gpt-4-vision?
I tried replacing the “headers” in response = client.chat.completions.create(). However, I am still getting an error.
Welcome to the forum.
I’ve heard that some domains are blocking it… Can you host on your own or another server?
I would prefer to directly query them to have a stable pipeline, however I would also look into hosting it myself if it really is impossible to query directly. Thanks.
Yeah, you have to think about it from a webmaster’s perspective. They might not enjoy a bot hitting their images dozens or hundreds or thousands of times and have to pay for bandwidth, etc. It’s not a ton, but it adds up. Could be the principle of it too.
I remember in the olden days with hotlinking images, some webmasters would put unsavory images to be served if request didn’t come from the original domain. I don’t think that happens much anymore, thankfully!
Yeah agreed. I also found it a bit faster to just use local storage to query a bunch of images at the same time rather than using a web scraper for it. Thanks for the suggestion.
1 Like