Why do you need some groundtruth values for your reasoning/testing?

[Rango-T10000]

According to your forward_test code, it still needs some groundtruth values as input. And your data info provides these groundtruth values. Why do you still need some groundtruth values for your reasoning/testing? For inference, Shouldn't input data just count pictures?

[Rango-T10000]

May I ask is to calculate the evaluation index after reasoning

[BaophanN]

Hi, have you been able to design the inference script? Thank you