Efficient Neighbourhood Consensus Networks via Submanifold Sparse Convolutions
本文的目的是输入一个image pair 然后得到他们的匹配 内存消耗大，推理时间长，对应关系局部性差。我们提出的修改可以减少10倍以上的内存占用和执行时间，并且效果相当。这是通过对包含试探性匹配的相关张量进行稀疏化，然后使用子流形稀疏卷积对其进行4D CNN后续处理来实现的。通过以更高的分辨率处理输入图像（这是可能的，因为减少了内存占用），以及通过一个新的两级对应重定位模块，定位精度显著提高
相关文章
- 1. Neighbourhood Consensus Networks
- 2. 【文章阅读】Submanifold Sparse Convolutional Networks
- 3. 【稀疏卷积】Submanifold Sparse Convolutional Networks
- 4. 论文笔记-IGCV3：Interleaved Low-Rank Group Convolutions for Efficient Deep Neural Networks
- 5. 论文笔记：An Efficient Hardware Accelerator for Structured Sparse Convolutional Neural Networks on FPGAs
- 6. 论文笔记：An Efficient Hardware Accelerator for Sparse Convolutional Neural Networks on FPGAs
- 7. Robust Face Recognition via Sparse Representation
- 8. Efficient Neural Architecture Search via Parameter Sharing
- 9. Efficient Video Object Segmentation via Network Modulation
- 10. MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
- 更多相关文章...
- • PHP is_uploaded_file() 函数 - PHP参考手册
- • Docker Compose - Docker教程
相关标签/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
欢迎关注本站公众号,获取更多信息
相关文章
- 1. Neighbourhood Consensus Networks
- 2. 【文章阅读】Submanifold Sparse Convolutional Networks
- 3. 【稀疏卷积】Submanifold Sparse Convolutional Networks
- 4. 论文笔记-IGCV3：Interleaved Low-Rank Group Convolutions for Efficient Deep Neural Networks
- 5. 论文笔记：An Efficient Hardware Accelerator for Structured Sparse Convolutional Neural Networks on FPGAs
- 6. 论文笔记：An Efficient Hardware Accelerator for Sparse Convolutional Neural Networks on FPGAs
- 7. Robust Face Recognition via Sparse Representation
- 8. Efficient Neural Architecture Search via Parameter Sharing
- 9. Efficient Video Object Segmentation via Network Modulation
- 10. MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications