Introduction

© Haodong Li haodongli@zju.edu.cn

© Zhengnan Sun 3190102989@zju.edu.cn

Basic Information

  • Soft Machine Vision
  • Smart Factory Innovation Club of Zhejiang University
  • Class locations: 紫金港月牙楼301
  • Class time: Sunday 9:30 to 11:30
  • Number of students enrolled: 42

Course Outline

  • Image basics: Pixels, Colors, Image formats
  • Image processing techniques: Filtering, Binarization, cutting, Morphological transformation, Scale and rotation transformation, Image gradient
  • Image pattern recognition: Line and circle detection, Feature point detection, Edge detection
  • Image pattern recognition: Blob detection, Feature point detection, Template matching
  • Neural Network basics: Neuron structure, Multi-layer Perceptron, Handwritten digit recognition, Loss function
  • Neural Network basics: Gradient Descent, Backpropagation
  • Modern Neural Network: Softmax regression, Deep neural network, Convolutional Neural Network, Server Resources
  • Modern Neural Network: Recurrent Neural Networks, Attention Mechanism & Transformer, Natural Language Processing, Reinforcement Learning, Generative Adversarial Networks
  • Final project and Q&A: Chess board recognition system

Configuration (pip)

  • Python3.9 [LINK]
  • Add Python to PATH (environment variables)
  • Install required packages
    • For CPU users
      • pip install -r ./0-introduction/requirements_cpu.txt
    • For GPU users (please make sure CUDA 11.3 and cuDNN 8.2 are installed in your computer)
      • pip install -r ./0-introduction/requirements_gpu.txt
  • Please be aware of the dependencies if you have more than one Python installed
  • VS Code [LINK] (optional)
    • Config Python3.9 in your VS Code
  • Miniconda installation package download [LINK]
  • Add Miniconda to PATH (environment variables)
  • Open terminal in Vision2022/
  • Create a virtual environment (replace <environment name> with a name given by you, e.g. vision)
    • conda create -n <environment name> python=3.9
  • Activate the virtual environment
    • conda activate <environment name>
  • Install required packages
    • For CPU users
      • pip install -r ./0-introduction/requirements_cpu.txt
    • For GPU users (cudatoolkit will config CUDA 11.3 and cuDNN 8.2 automatically, so you don’t need to install them manually)
      • conda install pytorch==1.10.2 torchvision==0.11.3 torchaudio==0.10.2 cudatoolkit=11.3 -c pytorch -c conda-forge
      • pip install -r ./0-introduction/requirements_gpu_conda.txt

Test configuration

  • Run python ./0-introduction/show_a_cat.py under Vision2022/
    • Press Esc or q on the keyboard to close the window
import sys

print(sys.version)

3.9.16 (main, Mar  8 2023, 10:39:24) [MSC v.1916 64 bit (AMD64)]

import torch
import tensorflow as tf

print(torch.cuda.is_available())
# let's see the list of CUDA architectures, and the device name
if torch.cuda.is_available():
    print(torch.cuda.get_device_name(device=None), torch.cuda.get_arch_list())
print(tf.test.is_gpu_available())
# tf.test.is_gpu_available() is deprecated and will be removed soon
tf.config.list_physical_devices('GPU') 

True
NVIDIA GeForce RTX 3060 Laptop GPU ['sm_37', 'sm_50', 'sm_60', 'sm_61', 'sm_70', 'sm_75', 'sm_80', 'sm_86', 'compute_37']
WARNING:tensorflow:From C:\Users\lebro\AppData\Local\Temp\ipykernel_19472\519669526.py:8: is_gpu_available (from tensorflow.python.framework.test_util) is deprecated and will be removed in a future version.
Instructions for updating:
Use `tf.config.list_physical_devices('GPU')` instead.
True





[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]

import cv2
import numpy as np
import matplotlib.colors as mat_color

print(cv2.__version__)

# read the cat image
path = "./images/cat.jpg"
img_bgr = cv2.imread(path) 

print(type(img_bgr))
print(np.shape(img_bgr))

4.7.0
<class 'numpy.ndarray'>
(493, 493, 3)

from matplotlib import pyplot as plt

# display img
no_norm = mat_color.Normalize(vmin=0, vmax=255, clip=False)
plt.imshow(img_bgr, norm=no_norm)

<matplotlib.image.AxesImage at 0x7ff9561ede80>

png

# bgr -> rgb
img_rgb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)
plt.imshow(img_rgb, norm=no_norm)

<matplotlib.image.AxesImage at 0x7ff954150040>

png

import os

# save the img in csv format
os.makedirs('./data', exist_ok=True)
data_file = os.path.join('./data', 'cat.csv')
print(data_file)
with open(data_file, 'w') as f:
    f.write('R,G,B\n')
    for row in img_rgb:
        for rgb in row:
            f.write(str(rgb[0]) + ',' + str(rgb[1]) + ',' + str(rgb[2]) + '\n')

./data/cat.csv

The End

2022.3