Chuong Nguyen
Chuong Nguyen
Thanks for the opensource. Can we support 3DSSD https://github.com/Jia-Research-Lab/3DSSD? It seems achieve both good speed and high AP. Thank you.
@fellenB Hi, do you have any progress on adding Deep Closest Points, thanks?
No, taking the limit when dS/dA converges to 0, alpha converges to 0. Otherwise, you can divide it.
>Hi! Does this mean the class-agnostic detection score is not carried to the 2nd stage at training? @xingyizhou Can you help to clarify this question ?
Hi @Ben-Louis @jshilong , I am trying your code with a Box_Head that has DW Conv. However, I got the error in this step: `module_ancest = set(self.conv2ancest[module]) ` or `module_ancest...
I am curiousl too, since the original paper proposed the Micro-block as: BN-Act-Conv2d. This is also replicated by other pytorch implementation, and also used in DenseNet. However, the ResNeXt uses...
Hi thanks Relh. Actually, in the paper "Identity mappings in deep residual networks", He et al. examined several structures, include BN-Act-Cov, and Cov-Bn-Act, and found that the structure BN-Act-Conv is...
"SyncBN" is only for Multi-GPU. Set it to BN or GN for single GPU
@Ezra-Yu Interesting, Do we have the speed comparision between different implements yet?
@yangcf10 @yuweihao >I understand it's mostly from empirical results. But any insight why we should do the subtraction? Average Pool combining with Subtraction yields a [Laplacian kernel] (https://homepages.inf.ed.ac.uk/rbf/HIPR2/log.htm) [1 1...