<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[adaadadsa的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/adaadadsa</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; adaadadsa]]></copyright><item><title><![CDATA[Linear regression pytorch]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/132265672</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/132265672</guid><author>adaadadsa</author><pubDate>Sun, 13 Aug 2023 21:25:39 +0800</pubDate><description><![CDATA[true_b = 3epoch = 3lr = 0.03loss = 0。]]></description><category></category></item><item><title><![CDATA[deep learning linear regression demo1]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/132248516</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/132248516</guid><author>adaadadsa</author><pubDate>Sat, 12 Aug 2023 14:28:02 +0800</pubDate><description><![CDATA[【代码】deep learning linear regression demo1。]]></description><category></category></item><item><title><![CDATA[Numpy实现简单的线性回归]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/132140315</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/132140315</guid><author>adaadadsa</author><pubDate>Mon, 07 Aug 2023 09:53:36 +0800</pubDate><description><![CDATA[【代码】Numpy实现简单的线性回归。]]></description><category></category></item><item><title><![CDATA[多项式求和算法对比]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/126915038</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/126915038</guid><author>adaadadsa</author><pubDate>Sun, 18 Sep 2022 10:56:09 +0800</pubDate><description><![CDATA[多项式求和算法对比]]></description><category></category></item><item><title><![CDATA[UserWarning: Corrupt EXIF data. Expecting to read 4 bytes but only got 0. warnings.warn(str(msg)]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/125818329</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/125818329</guid><author>adaadadsa</author><pubDate>Sat, 16 Jul 2022 12:19:07 +0800</pubDate><description><![CDATA[图片加载出错，删除文件]]></description><category></category></item><item><title><![CDATA[Python 库]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/124491966</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/124491966</guid><author>adaadadsa</author><pubDate>Fri, 29 Apr 2022 11:22:51 +0800</pubDate><description><![CDATA[https://www.lfd.uci.edu/~gohlke/pythonlibs/



numpy:https://www.lfd.uci.edu/~gohlke/pythonlibs/#numpy

pytorch:https://www.lfd.uci.edu/~gohlke/pythonlibs/#pytorch
]]></description><category></category></item><item><title><![CDATA[conda 安装第三方包]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/118853482</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/118853482</guid><author>adaadadsa</author><pubDate>Sat, 17 Jul 2021 10:45:10 +0800</pubDate><description><![CDATA[下载whl包到本地：https://pypi.org/project/pygame/#files

激活环境 activate xxxx；

pip install path\whl
]]></description><category></category></item><item><title><![CDATA[torch.sum()]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/108426001</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/108426001</guid><author>adaadadsa</author><pubDate>Sat, 05 Sep 2020 23:12:04 +0800</pubDate><description><![CDATA[import torch

a = torch.randint(0,3,(2,3,4))
print(a.size())
print(a)
b = torch.sum(a, 0) # 按将第一个纬度的值相加
c = torch.sum(a, 1) # 按将第二个纬度的值相加
print('b', b)
print('c', c)


]]></description><category></category></item><item><title><![CDATA[PIL转tensor再传PIL并打印图片]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/108247846</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/108247846</guid><author>adaadadsa</author><pubDate>Wed, 26 Aug 2020 21:19:15 +0800</pubDate><description><![CDATA[import torch.nn as nn
from torchvision import transforms
from PIL import Image
import torch


img = Image.open('../data/pikachu.jpg')

transform_to_tensor = transforms.ToTensor()
img_ = transform_to_tensor(img)

print(1, img_.size())

img_ = torch.unsquee.]]></description><category></category></item><item><title><![CDATA[yolo v1 loss]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/108049383</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/108049383</guid><author>adaadadsa</author><pubDate>Mon, 17 Aug 2020 09:41:16 +0800</pubDate><description><![CDATA[yolov1 loss学习模型输出解析7x7x30特征图并计算loss
模型输出
yoloV1模型最后连接两个全连接层，分别把特征图尺寸归一到4096和想要的7x7x30，怎样对7x7x30进行解析并计算loss的呢
解析7x7x30特征图并计算loss
模型输出：


模型输出会有很多个batch，每个batch输出就像上图所示，假设为output，为7x7x30的特征图，所以对每个batch进行遍历，得到当前batch特征图，然后对每个grid进行遍历，即遍历特征图的宽和高，得到output[bi, w]]></description><category></category></item><item><title><![CDATA[卷积核卷积深度和个数的理解]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/108034650</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/108034650</guid><author>adaadadsa</author><pubDate>Sun, 16 Aug 2020 12:26:56 +0800</pubDate><description><![CDATA[卷积运算为输入的特征图input与卷积核kernel进行卷积运算，input 尺寸一般为batch*channel*width*high,然后卷积核我们一般只说两个尺寸,width * high，其实卷积核也有深度，跟input的channel数一样，然后有卷积个数，即输入特征图的个数。

每次运算是，去一个batch的特征图，即channel*width*high简称c*w*h与卷积核kernel的w*h进行卷积运算，此处卷积核深度其实跟channel值一样，比如3*3*3的特征图与卷积核尺寸为3*3的卷]]></description><category></category></item><item><title><![CDATA[python PIL 等比例缩放图片，并将图片补充为正方形]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/107946390</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/107946390</guid><author>adaadadsa</author><pubDate>Tue, 11 Aug 2020 22:38:12 +0800</pubDate><description><![CDATA[from PIL import Image

img = Image.open('./testpic.jpg')
w, h = img.size
print(w, h)
w_s = int(w/2)  # 长宽缩小两倍
h_s = int(h/2)  # 长宽缩小两倍
img = img.resize((w_s, h_s),Image.ANTIALIAS)
blank = (w_s - h_s)/2

# img.crop((w0, h0, w1, h1)) w0，h0宽度，高度起始方向剪裁的值，为负时是.]]></description><category></category></item><item><title><![CDATA[LSTM学习笔记]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/107850190</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/107850190</guid><author>adaadadsa</author><pubDate>Thu, 06 Aug 2020 21:30:42 +0800</pubDate><description><![CDATA[学习自：https://blog.youkuaiyun.com/yingqubaifumei/article/details/100888147

这篇文章的动图真的不错。


]]></description><category></category></item><item><title><![CDATA[torch.split()方法]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/107804309</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/107804309</guid><author>adaadadsa</author><pubDate>Tue, 04 Aug 2020 23:51:51 +0800</pubDate><description><![CDATA[torch.split(tensor,split_size_or_sections,dim=0)

split_size_or_sections 为切分后的每块大小，不是切分为多少块


import torch

x = torch.randn(1, 2, 4, 4)
y = torch.split(x, 1, dim=1)    # 每块大小为1
# print(x[0])
for i in y:
    print(i.size())

a = torch.rand(1, 4, 8, 6)
b =..]]></description><category></category></item><item><title><![CDATA[python opencv 实现膨胀和腐蚀，开运算，闭运算，顶帽操作，黑帽操作]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/107500383</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/107500383</guid><author>adaadadsa</author><pubDate>Tue, 21 Jul 2020 22:22:01 +0800</pubDate><description><![CDATA[import cv2
import numpy as np


# 腐蚀kernel
dilate_kernel = np.ones((3,3))
# 膨胀kernel
erose_kenel = np.array([[255,255,255],[255,255,255],[255,255,255]])

test_img = cv2.imread('../data/testpic.jpg')
print(test_img.shape)

# 膨胀或腐蚀的迭代次数
test_img1 = cv2.dila.]]></description><category></category></item><item><title><![CDATA[感受野计算公式]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/107442242</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/107442242</guid><author>adaadadsa</author><pubDate>Sun, 19 Jul 2020 11:02:35 +0800</pubDate><description><![CDATA[# 卷积计算公式 D_in：输入特征图尺寸， p：padding大小，k：kernel_size卷积核大小，
# s:stride 步距，D_out ：输出特征图尺寸
D_out = (D_in + 2p - k) / s + 1

# 反过来则得到感受野计算公式
D_in = (D_out - 1) * s + k - 2p

"""
例 5*5 的特征图经过 3*3 的卷积核，s为1，进行卷积运算得到 3*3的特征图，

那么3*3的特征图在3*3的卷积核，s为1情况下，感受野为 5*5

"""

.]]></description><category></category></item><item><title><![CDATA[tensor图像的显示]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/107301661</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/107301661</guid><author>adaadadsa</author><pubDate>Sun, 12 Jul 2020 16:51:49 +0800</pubDate><description><![CDATA[from PIL import Image
import matplotlib.pyplot as plt
import torch
import torchvision.transforms as transforms

ori_img = Image.open('./dog1.jpg')
# plt.imshow(ori_img)
# plt.show()
print(ori_img.size)
my_transforms = transforms.Compose([
    transforms.R.]]></description><category></category></item><item><title><![CDATA[AdaptiveAvgPool2d 测试]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/107030224</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/107030224</guid><author>adaadadsa</author><pubDate>Mon, 29 Jun 2020 22:43:36 +0800</pubDate><description><![CDATA[import torch.nn as nn
import torch


# 创建三维tensor
a = torch.randn(3,4,5)
print(a.shape)
print(a)
# 升维，升成四维
a = torch.unsqueeze(a, 0)
print(a.shape)
print(a)
# AdaptiveAvgPool2d(X) 是将W H 使用平均池化降为X维
avg = nn.AdaptiveAvgPool2d(1)
b = avg(a)
print(a.shape)
pr.]]></description><category></category></item><item><title><![CDATA[plt显示灰度图出现Invalid shape问题]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/106984133</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/106984133</guid><author>adaadadsa</author><pubDate>Sat, 27 Jun 2020 16:17:08 +0800</pubDate><description><![CDATA[转自：https://blog.youkuaiyun.com/grllery/article/details/102470263


transform = transforms.Compose([
    transforms.Grayscale(num_output_channels=1),
    transforms.ToTensor()
])

img = transform(img) 
print(img.size()) # img(1, 681, 1023)
img = np.transpose(img,]]></description><category></category></item><item><title><![CDATA[pytorch彩色图像转灰度图像]]></title><link>https://blog.youkuaiyun.com/adaadadsa/article/details/106983918</link><guid>https://blog.youkuaiyun.com/adaadadsa/article/details/106983918</guid><author>adaadadsa</author><pubDate>Sat, 27 Jun 2020 16:03:49 +0800</pubDate><description><![CDATA[from torchvision import transforms



transform = transforms.Compose([
    transforms.Grayscale(num_output_channels=1), # 彩色图像转灰度图像num_output_channels默认1
    transforms.ToTensor()
])


]]></description><category></category></item></channel></rss>