search

zhaoweicai / mscnn

Caffe implementation of our multi-scale object detection framework
404 stars 211 forks source link

what's the context embedding? #76

Open OxInsky opened 7 years ago

OxInsky commented 7 years ago

It's my pleasure to read your paper.there is a question to me in this paper. Would you explaining the context embedding to me for further details ? how do you do it? Look forward to your reply！

zhaoweicai commented 7 years ago

The details can be found in Section 4.2 and Fig. 4 in the paper. ROI pooling is on two regions (object region and context region). The outputs from two ROI Pooling operations are stacked together. And then a convolution follows the stacked ROI Pooling outputs directly.