Skip to content

Commit

Permalink
Update file
Browse files Browse the repository at this point in the history
  • Loading branch information
HarleysZhang committed Oct 9, 2021
1 parent e478b63 commit 40c7016
Show file tree
Hide file tree
Showing 6 changed files with 32 additions and 0 deletions.
20 changes: 20 additions & 0 deletions .vscode/launch.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,20 @@
{
// Use IntelliSense to learn about possible attributes.
// Hover to view descriptions of existing attributes.
// For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
"version": "0.2.0",
"configurations": [
{
"name": "cl.exe - 生成和调试活动文件",
"type": "cppvsdbg",
"request": "launch",
"program": "${fileDirname}\\${fileBasenameNoExtension}.exe",
"args": [],
"stopAtEntry": false,
"cwd": "${fileDirname}",
"environment": [],
"console": "externalTerminal",
"preLaunchTask": "C/C++: cl.exe 生成活动文件"
}
]
}
4 changes: 4 additions & 0 deletions 5-deep_learning/4-深度学习面试题.md
Original file line number Diff line number Diff line change
Expand Up @@ -139,6 +139,10 @@

+ 举例理解就是`Pytorch` 的卷积层和池化层的输入 `shape` 格式为 `(N, C, H, W)``Keras` 的卷积层和池化层的输入 `shape` 格式为 `(N, H, W, C)`

值得注意的是 `OpenCV` 读取图像后返回的矩阵 `shape` 的格式是 `(H, W, C)`格式。当 OpenCV 读取的图像为彩色图像时,返回的多通道的 BGR 格式的矩阵(`HWC`),在内存中的存储如下图:

![opencv矩阵存储格式](../data/images/opencv/opencv图像矩阵的存储格式.png)

### 四,Pytorch 、Keras 的池化层函数理解

> 注意:对于 `Pytorch、Keras` 的卷积层和池化层函数,其 **`padding` 参数值都默认为不填充像素**,默认值为 `0``valid`
Expand Down
5 changes: 5 additions & 0 deletions 6-computer_vision/目标检测/3-Mask RCNN理解.md
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
- [ROI Pooling 和 ROI Align 的区别](#roi-pooling-和-roi-align-的区别)
- [Mask R-CNN 网络结构](#mask-r-cnn-网络结构)
- [骨干网络 FPN](#骨干网络-fpn)
- [anchor 锚框生成规则](#anchor-锚框生成规则)
Expand All @@ -6,6 +7,10 @@

> `Mask RCNN` 是作者 `Kaiming He``2018` 年发表的论文
## ROI Pooling 和 ROI Align 的区别

[Understanding Region of Interest — (RoI Align and RoI Warp)](https://towardsdatascience.com/understanding-region-of-interest-part-2-roi-align-and-roi-warp-f795196fc193)

## Mask R-CNN 网络结构

`Mask RCNN` 继承自 `Faster RCNN` 主要有三个改进:
Expand Down
2 changes: 2 additions & 0 deletions 6-computer_vision/目标检测/4-RetinaNet网络理解.md
Original file line number Diff line number Diff line change
Expand Up @@ -80,6 +80,8 @@

### 4.1,Cross Entropy

> 可额外阅读文章 [理解交叉熵损失函数](https://zhuanlan.zhihu.com/p/339684056)
在深度学习中我们常使用交叉熵来作为分类任务中训练数据分布和模型预测结果分布间的代价函数。对于同一个离散型随机变量 $\textrm{x}$ 有两个单独的概率分布 $P(x)$ 和 $Q(x)$,其交叉熵定义为:
> P 表示真实分布, Q 表示预测分布。
Expand Down
1 change: 1 addition & 0 deletions 7-model_deploy/3-模型量化.md
Original file line number Diff line number Diff line change
Expand Up @@ -277,6 +277,7 @@ TensorRT 使用 KL 散度算法进行量化校准的过程:首先在校准集

## 参考资料

- [NCNN Conv量化详解(一)](https://zhuanlan.zhihu.com/p/71881443)
- [卷积神经网络优化算法](https://jackwish.net/2019/convolution-neural-networks-optimization.html)
- [神经网络量化简介](https://jackwish.net/2019/neural-network-quantization-introduction-chn.html)
- [QNNPACK 实现揭秘](https://jackwish.net/2019/reveal-qnnpack-implementation.html)
Expand Down
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.

0 comments on commit 40c7016

Please sign in to comment.