@zd11024 Thanks for your great works！
Using custom data from a living room scene for model inference yields poor results. I think vgllm have zero-shot capabilities for living room scenes, as the training data includes many living room scene samples. Do you have any suggestions?
Thanks！
@zd11024 Thanks for your great works！
Using custom data from a living room scene for model inference yields poor results. I think vgllm have zero-shot capabilities for living room scenes, as the training data includes many living room scene samples. Do you have any suggestions?
Thanks！