Poor performance of the other question.

[ZebinX]

Hi, @gordonhu608

Thanks for great work.

I tried inference_chat.py to ask some other questions, like the comparison of the locations, but G2VLM seems could not answer properly. I'm not sure if I use model correctly, could you help me figure it? My case:

Fig:

Question:
which one is closer to me? The blue chair or the white shoes
Answer:
The observer is 1.4 meters away from the center of shoes (red point)

It seems can only answer the question about the depth.

Thank you!