While using the pretrained model, I've observed that many queries in pred_logits are predicting class 1, which corresponds to cars. For each output I examined, a significant number of queries seem to be identifying class 1. Is this behavior due to the prevalence of multiple cars in a single frame of dataset, leading multiple queries to recognize different cars?
Hello,
While using the pretrained model, I've observed that many queries in pred_logits are predicting class 1, which corresponds to cars. For each output I examined, a significant number of queries seem to be identifying class 1. Is this behavior due to the prevalence of multiple cars in a single frame of dataset, leading multiple queries to recognize different cars?
Thanks