- 博客(4)
- 收藏
- 关注
RSS订阅原创 VoteNet学习笔记
传统2D霍夫投票onlineFirst, given a collection of images with annotated object bounding boxes, a codebook is constructed with stored mappings between image patches (or their features) and their offsets to the corresponding object centers.用image和gt bbox构建一个
2021-05-24 17:16:57
1377
1
原创 PointRCNN阅读笔记
Instead of generating proposals from RGB image or projecting point cloud to bird's view or voxels as previous methods do, our stage-1 sub-network directly generates a small number of high-quality 3D proposals from point cloud in a bottom-up manner via seg.
2021-05-24 10:25:14
399
原创 spconv学习笔记
SparseConvolution__init__输入：ndim：kernel_size, stride等需要复制的维度in_channels: weight输入维度out_channels: weight输出维度kernel_size, stride, padding, dilation, groups, bias与Conv2d定义相同subm: bool, 有没有submanifoldoutput_padding: deconv是output需要padding的大小transpose
2021-05-14 20:30:22
2401
1
翻译 [翻译]Fast Adjustable Threshold for Uniform Neural Network Quantization
摘要提出了两种方法来优化量化训练：1. 引入离散化阈值的训练尺度因子，对每个滤波器进行分离。2. 后续的的dws卷积和卷积层的互相重标度。根据以上技术，本文用10%ImageNet2012的训练数据对移动网络进行了训练量化。这样减少了训练数据集大小和训练参数量，使得在保持量化模型高精度（少于0.5%的精度下降）的同时仅用几个小时就微调了网络。可用的模型和代码：https://github.co...
2019-09-25 14:05:33
289
空空如也
空空如也
TA创建的收藏夹 TA关注的收藏夹
TA关注的人