Visual Question Answering
Transformers
Safetensors
English
Chinese
minicpmv
feature-extraction
custom_code

Conditional chatting

#10
by plo33 - opened

Hi guys,

Can i use this model to have an efficient conditional chatting like:

  • If there is a red car OR black car OR yellow car in the image give me some details of image, otherwise, just answer 'NO'.

I have tested some phrases like this, but 99% of time it gives me 'NO', even if the image have the colored cars mentioned in the prompt.

OpenBMB org

I think you can make it into 2 stages.

  1. yes or no
  2. if yes then captioning

Sign up or log in to comment