I tested the pretrained weights to see how well the model generalizes. I used samples from Market1501 and pretrained weights from the MSMT17 dataset (bot, R50). The results showed decent similarity values. While the trends are acceptable, shouldn't the similarity be lower for different people?
I tested the pretrained weights to see how well the model generalizes. I used samples from Market1501 and pretrained weights from the MSMT17 dataset (bot, R50). The results showed decent similarity values. While the trends are acceptable, shouldn't the similarity be lower for different people?
Similarity calculated by using the below code: