why the last layer of retina_head is nn.con2d() of 3*3 ,not is nn.linear() , what difference are both of them?? - Githubissues

toandaominh1997 / EfficientDet.Pytorch

Implementation EfficientDet: Scalable and Efficient Object Detection in PyTorch

MIT License

1.44k stars 305 forks source link

why the last layer of retina_head is nn.con2d() of 3*3 ,not is nn.linear() , what difference are both of them?? #128

Open y78h11b09 opened 4 years ago