Plzz

Local Environment building(GPU Version) My environment version conda cuda cudnn cudatoolkit torch torchaudio torchvision Visual Studio 2022 4070TS 16G显存 How to SETUP Way 1 Download the .yaml file locally from the Deeplabcut home page 12cd Path_to_your_yaml_fileconda env create -f DEEPLABCUT.yaml If errors occur, you can use 1conda env update -f DEEPLABCUT.yaml to continue to create your env or update your env If the cuda version is still CPU, then you need to install pytorch gpu versi ...

Fitting Least Squares 最小二乘法 Assumption：仅考虑真实值与预测值之间的误差, 计算的是 y^\hat{y}y^​ 与 yyy之间的距离 Formula: E=argmina,b∑i=1N∥yi−axi−b∥2=∥Y−XB∥2E = argmin_{a,b}\sum_{i = 1}^N \|y_i - ax_i -b\|^2 = \|Y - XB \|^2E=argmina,b​∑i=1N​∥yi​−axi​−b∥2=∥Y−XB∥2 Solution: B=(XTX)−1XTYB = (X^TX)^{-1}X^TYB=(XTX)−1XTY 但是如果line是vertical，最小二乘法将会失效 Total Least Squares 总体最小二乘法 Assumption: 同时考虑因变量和自变量中的误差，使用直线表示法，可处理vertical case，计算的是 (xi,yi)(x_i,y_i)(xi​,yi​) 到直线 ax+by=dax+by =dax+by=d 的垂直距离 Formula: E=argmina,b,d∑i=1N∥axi+ ...

Point Cloud Introduction Intro: 点云是一种用于表示三维空间中对象或场景的数据结构。在原始的实现中，它其实是一个包含多个三维坐标点的集合(这些点是通过对实际场景或物体的表面进行离散采样而得到的)。 Def: 点云是场景表面在给定坐标系下的离散表示 Properties: 高度灵活性Flexibility: 可用于表示各种类型的三维形状 数据密度可变性: 可sparse, 可dense 附加信息丰富性: 除了坐标外，可含其他属性信息。比如在重建中，点云一般还携带着color/intensity value, time, etc的信息。 Package Requirement open3d in Python One thing need to focus is that your open3d version should be correct, otherwise it will lead to dependency problems. You can deal with this by lowing the open3d version. ...

Related Paper Time Efficient Ultrasound Localization Microscopy Based on A Novel Radial Basis Function 2D Interpolation Increasing frame rate of echocardiography based on a novel 2D spatio-temporal meshless interpolation Environment Building Matlab Install required toolboxs, including Communications, Bioinformatics, Image Processing, Curve Fitting, Signal Processing, Statistics and Machine Learning,Parallel Computing, and Computer Vision Toolbox. Run PALA_SetUpPaths.m to check whether you hav ...

Local Environment building What you need Hardware 可用CUDA的设备(因此MACOS不能run 3DGS) 足够的显存(最好24G) Software Conda C++ Compiler for Pytorch extensions(eg. Visual Studio) CUDA SDK 11 for Pytorch extensions 所有的环境要版本适配 My environment version conda 4.12.0 cuda 11.7 torch 1.12.1 torchaudio 0.12.1 torchvision 0.13.1 Visual Studio 2022 8G显存 How to SETUP 123SET DISTUTILS_USE_SDK=1 # Windows onlyconda env create --file environment.ymlconda activate gaussian_splatting Execute the above command, then it ca ...

DIP_quiz_3 - Image segmentation Q1 Thresholding Factors affecting thresholding Noise →\rightarrow→ Distribution of image, hard to execute thresholding process foreground ratio →\rightarrow→ Distribution of image, hard to execute thresholding process The uniformity of the illumination source →\rightarrow→ Distribution of image, hard to execute thresholding process Otsu’s method 唯一需要注意的点在于，新的m1与m2的计算，是需要重新计算组内的概率分布!!! 即P1∗m1+P2∗m2=mGP_1*m_1 + P_2*m_2 = m_GP1​∗m1​+P2​∗m2​=mG​, 而不是m1+m2=mGm_1 + ...

I also handwrote some details on the printout chceatsheet, showing as follows

DIP_quiz_2 - Transform domain processing Q1 Circular convolution Difference between linear(full) convolution, circular convolution and periodical convolution Linear convolution: the final length will be equal to M+N-1 and directly use the multiplication formula Circular convolution: the final length will be equal to the original length L use linear convolution to simulate by removing the addition lower bits to the target higher ones the conv item can be seen as the periodic signal and t ...