Hi, thank you for your nice work. I'm reading it recently but I got some questions about the paper. I would appreciate it if I could get your help with it. So, is "identity-aware useful information" that is implicit in the modality-specific features truly useful? As they just exist in specific modalities, they can't be shared features that we need in person-reid. If you want to search person across modalities, you should use shared features rather than unshared features. If a feature is already a modality-specific feature, how can you transfer it into a modality-shared feature? Maybe I didn't understand it right in some way. Could you explain it more specifically to me? Thank you.
Hi, thank you for your nice work. I'm reading it recently but I got some questions about the paper. I would appreciate it if I could get your help with it. So, is "identity-aware useful information" that is implicit in the modality-specific features truly useful? As they just exist in specific modalities, they can't be shared features that we need in person-reid. If you want to search person across modalities, you should use shared features rather than unshared features. If a feature is already a modality-specific feature, how can you transfer it into a modality-shared feature? Maybe I didn't understand it right in some way. Could you explain it more specifically to me? Thank you.