search

zhanggang001 / HEDNet

HEDNet (NeurIPS 2023) & SAFDNet (CVPR 2024 Oral)
Apache License 2.0
88 stars 7 forks source link

Why is it faster? #21

Open yinheyanxian opened 1 week ago

yinheyanxian commented 1 week ago

Hello, thank you for the open source code.

After reading your article and code, I have a question. Why is it faster on the Argo2 dataset than VoxelNeXt?

SAFDNet has significantly more Sparse convolutions and Sparse residual blocks than VoxelNeXt, which leads to better results, but why is it faster?

zhanggang001 commented 6 days ago

Hello, thank you for the open source code.

After reading your article and code, I have a question. Why is it faster on the Argo2 dataset than VoxelNeXt?

SAFDNet has significantly more Sparse convolutions and Sparse residual blocks than VoxelNeXt, which leads to better results, but why is it faster?

VoxelNeXt is faster than SFADNet (Table 5 of the paper). We do not claim that SAFDNet is faster than VoxelNeXt.