Closed ajay1234567899 closed 2 years ago
Please refer to our Arxiv paper, it includes all the detail on which data we used to pre-train and progressively finetune. In particular, have a look at section 7.1 to see how we obtain 7.5 MR-2 on citypersons.
Hello sir, Your repo mentioned that you had achieved state-of-the-art on the city person dataset by training on the extra train data, around 19k images.
I am not able to find the ground truth annotations of the same. Please do point out the location of the extra annotations.
Also, to get the state-of-the-art results of 7.5 MR-2 on the city persons you had used a training strategy, I cannot get the strategy properly. Please do point me to your pre-print version, which had its details Your current version on arxiv does not have sections 6.1 and 6.2