search

ZebangCheng / Emotion-LLaMA

Emotion-LLaMA: Multimodal Emotion Recognition and Reasoning with Instruction Tuning
BSD 3-Clause "New" or "Revised" License
103 stars 10 forks source link

About the prompts used in the paper #13

Open Arthurwyf opened 1 month ago

Arthurwyf commented 1 month ago

In the paper and readme, the prompts you used for minigpt and llama to generated desriptions is not illustrated in detail, cound you please give some examples?

ZebangCheng commented 1 month ago

The prompts of MiniGPT-v2:

"What do you see in the image? Please describe it using words only.",
"Please describe the content of the image.",

The prompts of LLaMA-3:

System: You are an emotion analysis expert. Please infer emotion label based on the given the emotional features.

Question: The woman in the video is talking to a man, possibly discussing something important or sharing her thoughts and feelings. The person’s expression and action include eyes widened, wide-mouthed, and speaking with a happy voice, saying: “Oh my.” Please sort out the correct emotional clues and infer why the person in the video feels surprise.