Skip to content
This repository has been archived by the owner on Sep 29, 2023. It is now read-only.

Base network selection #47

Open
DonaldKam opened this issue Mar 29, 2020 · 3 comments
Open

Base network selection #47

DonaldKam opened this issue Mar 29, 2020 · 3 comments

Comments

@DonaldKam
Copy link

As there are more popular and high-efficient base networks, i.e. ResNetX, why select VGG as the feature extractor? As we know, it enjoys tremendous parameters that slow down the computation speed and requires increasing memory.

@shijieS
Copy link
Owner

shijieS commented Apr 3, 2020

Yes, the feature extractor can be ResNet, DenseNet, etc. Actually, this code is based on SSD whose feature extractor is VGG. Also, you can try other feature extractors. I think you can get better results. Besides, we only employ the center of the box which is also the weakness of this network. There are lots of room for improvement.

@DonaldKam
Copy link
Author

Thank you and you remind me of another question. What is the center of the box? Coordinate? If not, will that be a patch cropped from a bbox?

@shijieS
Copy link
Owner

shijieS commented Apr 4, 2020

Features at the center of the box in the corresponding feature map.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants