search

jiasenlu / vilbert_beta

470 stars 96 forks source link

how do you extract the feature of the whole image by Faster-RCNN? #45

Open lowoodz opened 4 years ago

lowoodz commented 4 years ago

Hi, @jiasenlu jiasenlu, I am new to object recognition, I have read your paper and, it's mentioned that you have used whole image as a box and extracted its feature by Faster-RCNN. How do you achieve that? The RPN and Fast-RCNN are built into one caffe net model, do you use a customized Faster-RCNN? Can anyone help?

jiasenlu commented 4 years ago

Hi, I think I just average all the bounding box features to get the holistic image feature.

lowoodz commented 4 years ago

Thanks！

Regards Guanlin On 05/29/2020 12:30, Jiasen Lu wrote:

Hi, I think I just average all the bounding box features to get the holistic image feature.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.