fengbingchun/nn_test

This project contains some neural network code

NN_Test

This project contains some neural network code(Note: depend on opencv/libtorch/onnxruntime):

• neural network code(C++)
  • Perceptron
  • BP(Back Propagation)
  • CNN(Convolutional Neural Networks)
  • Linear Regression(gradient descent、least squares)
  • Naive Bayes Classifier(sex classification)
  • Logistic Regression(gradient descent, Batch/Mini Batch)
  • KNN(K-Nearest Neighbor, classification)
  • PCA(Principal Components Analysis)
  • single hidden layer(two categories)
  • K-Means
  • Decision Tree(CART)
  • YOLOv8(OpenCV DNN, libtorch, onnxruntime)
    • Object Detection
    • Instance Segmentation
• mathematical formula's implement with C++
  • Linear Algebra
    • transpose
    • determinant
    • adjoint matrix
    • inverse matrix
    • norm(vector、matrix)
    • eigenvalues/eigenvectors(real symmetric matrix)
    • SVD(Singular Value Decomposition)
    • pseudoinverse
    • trace
  • Probability
    • mean、variance、standard deviation
    • covariance matrix
  • Activation Function
    • logistic sigmoid function
    • softplus function
    • ReLU(rectified linear unit)
    • Leaky ReLUs(leaky rectified linear units)
    • ELUs(Exponential linear units)
    • softmax function
    • tanh(Hyperbolic tangent)
  • Loss Function
    • MSE(Mean Square Error)
    • Cross-Entropy
  • Optimization
    • BGD(Batch Gradient Descent)
    • SGD(Stochastic Gradient Descent)
    • MBGD(Mini-batch Gradient Descent)
    • SGD with Momentum
    • AdaGrad(Adaptive Gradient)
    • RMSProp(Root Mean Square Propagation)
    • Adadelta(an adaptive learning rate method)
    • Adam(Adaptive Moment Estimation)
    • AdaMax(a variant of Adam based on the infinity norm)
    • NAG(Nesterov Accelerated Gradient)
    • Nadam(Nesterov-accelerated Adaptive Moment Estimation)
  • Layers
    • Dropout
    • LRN(Local Response Normalization)
    • BN(Batch Normalization)
  • common algorithm
    • Brute Force(string match)
    • Triangle similarity(monocular ranging)
• open source library's usage(C++)
  • tiny-cnn GitHub
  • tiny-dnn GitHub
  • libsvm GitHub
  • OpenCV GitHub
    • Logistic Regression
    • Support Vector Machines(SVM)
    • KNN(K-Nearest Neighbor)
    • Decision Tree
    • PCA(Principal Components Analysis)
    • k-means
• Python code(support windows and linux platform)
  • Universal Configuration
    • Installation(Anaconda)

       conda create -n NN_Test python=3.9
       conda activate NN_Test
       pip install -r requirements.txt
       git clone https://github.com/fengbingchun/NN_Test
       cd NN_Test/demo/Python
      

    • Code
      • KNN(K-Nearest Neighbor)
      • Decision Tree(CART)
      • PCA(Principal Components Analysis)
      • k-means
        • k-means
        • k-menas opencv
      • txt to csv
  • MediaPipe
    • Installation(Anaconda)

       conda create -n MediaPipe python=3.9
       conda activate MediaPipe
       pip install mediapipe
       pip install requests
       git clone https://github.com/fengbingchun/NN_Test
       cd NN_Test/demo/Python
      

    • Code
      • face monocular ranging
  • Ultralytics
    • Installation(Anaconda)

       # cpu
       conda create --name ultralytics-env python=3.8 -y
       conda activate ultralytics-env
       pip install ultralytics
      
       # cuda
       # install cuda 11.8
       # install cudnn v8.9.7
       conda create --name ultralytics-env-cuda python=3.8 -y
       conda activate ultralytics-env-cuda
       conda install -c pytorch -c nvidia -c conda-forge pytorch==2.2.2 torchvision==0.17.2 torchaudio==2.2.2 pytorch-cuda=11.8 ultralytics # pytorch 2.2.2
      
       git clone https://github.com/fengbingchun/NN_Test
       cd NN_Test/demo/Python
      

    • Code
      • LabelMe JSON to YOLOv8 TXT: Detect
      • EISeg JSON to YOLOV8 TXT: Segment
      • YOLOv8 train
      • YOLOv8 predict
      • YOLOv8 segment onnxruntime predict
• Database convert to Images or Images conver to Database
  • MNIST
  • CIFAR10/CIFAR100
  • ORL Faces Database

The project support platform:

• windows10 64 bits: It can be directly build with VS2022 in windows10 64bits.
• Linux:
  • DatasetToImage support cmake build(file position: prj/linux_cmake_DatasetToImage)
  • Math support cmake build(file position: prj/linux_cmake_Math)
  • tiny-cnn support cmake build(file position: prj/linux_cmake_tiny-cnn)
  • tiny-dnn support cmake build(file position: prj/linux_cmake_tiny-dnn)
  • NN support cmake build(file position: prj/linux_cmake_NN)

Windows VS Screenshot:

Blog: fengbingchun

nn_test's Issues

这里pooling层也有w和b么，同时反卷积时卷积核为何没有翻转180

建议使用C++语法来实现类的结构

https://github.com/fengbingchun/NN_Test/blob/master/src/NN/CNN.hpp#L121 类中使用了大量的宏定义值magic number，夹杂使用了大量定长数组，以及若干指针变量。
既然类中也有使用vector容器，建议将这些C语言的特性以C++方式实现，就更加灵活易用了