MiniGPT-4 icon indicating copy to clipboard operation
MiniGPT-4 copied to clipboard

Published 20 hours ago • verified publisher icon Vision-CAIR

Does MiniGPT4 support multiple image uploading?

Open Leon1207 opened this issue 2 years ago • 3 comments

In some applications, you need to upload multiple pictures, clearly describe some scenes, and ask questions about the scenes. May I ask whether MiniGPT4 supports this practice?

Leon1207 avatar May 02 '23 09:05 Leon1207

I guess MiniGPT-4 can't support multiple images , but you can do these with a little bit extra works. Some suggestions I have tried maybe help you :

  1. use MiniGPT-4 to describe images as detailed as possible, try the best to get maximum information from the images
  2. use LLM as controller ,such as vincua、llama and so on, to finish your task. The information you get from images are the inputs of LLM, and you can prompt you question to LLM controller based on the inputs

feymanwang avatar May 06 '23 03:05 feymanwang

Thanks for your advice!

Leon1207 avatar May 06 '23 03:05 Leon1207

It can. In fact, I am drafting a PR to enable this.

ifsheldon avatar May 17 '23 09:05 ifsheldon

It can. In fact, I am drafting a PR to enable this.

hi,i think you will do a nice work,can you send me your work about supporting multiple images uploading to me?thanks!

wwcc1107 avatar Nov 15 '23 02:11 wwcc1107