About visual prompt.

[hengseuer]

Hi @Mountchicken ,

Suppose the batch size is set to 4.

In the first image, categories A and B have (N1_A) and (N1_B) instances, respectively. In the second image, categories C and D have (N2_C) and (N2_D) instances, respectively. In the third image, categories A and B have (N3_A) and (N3_B) instances, respectively. In the fourth image, categories C and D have (N4_C) and (N4_D) instances, respectively.

For category A in the first image, is its visual prompt selected randomly from 1 to N1, or from 1 to (N1 + N3)? If category C is the negative sample category for the first image, is its visual prompt selected randomly from 1 to N2 (N4), or from 1 to (N2 + N4)?

[Mountchicken]

Hi @hengseuer

• For category A in both the first and the third image, its visual prompt is selected randomly from 1 to (N1 + N3)
• In the first image, category C will be the negative prompts, sampled from 1 to (N2 + N4)

[hengseuer]

Thanks. Understood.