• Guide

• Define the “Industrial Vision Detection Brain”! The VisionBank AI 2.0 new version release conference was successfully held in Suzhou.

• What is the difference between machine vision and computer vision? 2022-06-16

• Human target detection car based on Raspberry Pi and YOLOv3 model (1)

• Human target detection car based on Raspberry Pi and YOLOv3 model (2)

• Human target detection car based on Raspberry Pi and YOLOv3 model (3)

• Human target detection car based on Raspberry Pi and YOLOv3 model (4)

• Build a car-mounted license plate recognition detection system purely by hand with Raspberry Pi, turning a family car into a smart car-G

• Machine vision innovation promotes automation forward-G

• Implementing basic machine vision technology with python+opencv (edge extraction, image filtering, edge detection operators, projection, license plate character segmentation)-G 2

• How to evaluate Google’s proposed MLP-Mixer: Can it achieve SOTA on ImageNet with just MLP?

• Google surpasses itself with new AI: allows Imagen to generate specified objects with arbitrary style conversion.

• “Defeating everyone in everything,” Microsoft’s multimodal new work sweeps 12 categories of tasks, even refreshing pure visual SOTA.

• A model solves common elements of the periodic table: A Chinese team creates a molecular simulation pre-training model that saves up to 90% of data.

• CVPR 2022 March 3 paper express (22 papers bundled for download) covers network architecture design, pose estimation, 3D vision, action detection, semantic segmentation, and other directions.

• Using computer vision to clean the world’s oceans 2022-02-23

• Improving visual systems: Multi-lens computational imaging provides enhanced capabilities for machine vision tasks 2021-12-26

• Visual Recruitment Bulletin Board

• Visual Project Supply and Demand Information

Guide

Define the “Industrial Vision Detection Brain”! The VisionBank AI 2.0 new version release conference was successfully held in Suzhou.

On August 27, 2022, the “Smart Manufacturing·VisionBank AI 2.0” new product launch conference was successfully held in Suzhou. After a year and a half of refinement, the company broke through bottlenecks and advanced to deeper AI technologies, launching the general intelligent vision development platform VisionBank AI 2.0, which will accelerate the industrial transformation of manufacturing enterprises.

At this launch conference, the company invited nearly a hundred representatives from manufacturing enterprises, universities, research institutes, and deep cooperation customers to discuss the new breakthroughs in technology of VisionBank AI 2.0 and explore the new trends in machine vision industry applications.

VisionBank AI 2.0: Continuing the integration of traditional algorithms and deep learning.

The realization of intelligent manufacturing cannot be separated from the production line + industrial vision. Using machine vision to deploy production line equipment to reduce costs and increase efficiency has become a consensus in the manufacturing industry, and advancing together and upgrading together is a major trend for future development. In the entire integrated system of machine vision, software and algorithms are the technical core of system realization.

……Long press the QR code to access the original text

The difference between machine vision and computer vision? 2022-06-16

Simple distinction:

Discipline:

Machine vision, as a system engineering discipline, is different from computer vision, which is a form based on computer science;

Computer vision belongs to computer science, involving the theory behind artificial systems that extract information from images, and it is interdisciplinary.

Machine vision is the application of computer vision in factory automation, and traditional machine vision is mainly applied in industrial fields, while computer vision is not limited to industrial fields.

Narrowly speaking, from the perspective of visual systems, computer vision is part of machine vision systems.

Information processing degree:

Machine vision mainly extracts information, while computer vision extracts and understands information (definitional distinction).

Hardware and software:

A machine vision system must contain hardware.

A computer vision system does not necessarily contain hardware, focusing on algorithms.

Details:

……Long press the QR code to access the original text

Human target detection car based on Raspberry Pi and YOLOv3 model (1) 2022-04-08

Project introduction:

The undergraduate graduation project chosen was deep learning. Initially, I only learned about the YOLOv3 model, downloaded the weights from the official website according to the author’s instructions, set up the environment, and ran the demo. Later, I thought that just running the model might be too simple, so I came up with the idea of creating a practical device (since I couldn’t optimize the model). Thus, the idea of making a controllable mobile car that can detect humans was born.

Functions implemented: 1. Control the car to move and detect human targets in real time.

2. As a home surveillance system, it can send the detected human targets appearing in the camera to a mobile phone via WeChat and allow the user to decide whether to trigger an alarm through a buzzer.

The general work includes: 1. Training the YOLOv3 tiny model.

2. Converting the Darknet model to TensorFlow model and then to NCS (Neural Compute Stick) model twice.

3. Car control and video streaming program.

4. WeChat alarm program.

1. Environment setup.

2. Install NVIDIA graphics driver.

……Long press the QR code to access the original text

Human target detection car based on Raspberry Pi and YOLOv3 model (2) 2022-04-08

The previous article introduced how to set up the deep learning environment on Ubuntu 18.04 TLS with CUDA: 9.0 + cuDNN 7.0 + tensorflow-gpu 1.9. This article will introduce how to create your own dataset and train the model.

The training dataset includes 6095 human images extracted from the VOC dataset and 200 human images obtained using the LabelImg tool as a supplement.

1. Crawling human images and labeling.

# coding=utf-8 “”” Download Baidu images based on search terms “”” import re import sys import urllib import requests def getPage ( keyword , page , n ): page = page * n keyword = urllib . parse

……Long press the QR code to access the original text

Human target detection car based on Raspberry Pi and YOLOv3 model (3) 2022-04-08

Model performance:

In the previous text, we created the dataset and trained the model using the dataset. We tested the detection performance of the model using static images and videos and found that the results were quite good.

The first two images are detection results from static images, and the last image is a screenshot of the detection results from a video.

However, to deploy the model on the power-limited Raspberry Pi, it needs to undergo two model conversions to run inference on the NCS.

Model conversion:

First conversion: (.weight –> .pb)

The official guide for this model conversion is provided by OpenVINO at https://docs.openvinotoolkit.org/latest/_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_YOLO_From_Tensorflow.html, but errors may occur.

python3 convert_weights_p

……Long press the QR code to access the original text

Human target detection car based on Raspberry Pi and YOLOv3 model (4) 2022-04-08

The previous articles completed the setup of the training and deployment environments and the training of the model, and after two model conversions, the YOLOv3 TINY model was finally deployed on the Raspberry Pi. In fact, the core content has been completed, and what follows are some application-level things.

Raspberry Pi controls the motor:

1. Hardware materials:

1. Raspberry Pi 3B+

2. Four DC motors

3. A car chassis + four wheels (available on Taobao)

4. L298N driver module (a bridge between the Raspberry Pi and the motors)

5. A power bank (for powering the Raspberry Pi)

6. Two 18650 lithium batteries (to power the car motors, ordinary dry batteries really can’t be used, otherwise you will have a lot of waste batteries)

7. 40 Dupont wires of male to male, female to female, and male to female

8. Breadboard (optional)

9. Buzzer

2. Circuit connections:

The image was drawn using the fritzing software, and it was my first time using it, so it looks a bit ugly, but it’s okay. I couldn’t find the L298N module, so I used the red one in the picture.

……Long press the QR code to access the original text

Build a car-mounted license plate recognition detection system purely by hand using Raspberry Pi, transforming a family car into a smart car-G 2022-04-06

Source: Machine Heart Translation Author: Robert Lucian Chiriac Out of boredom, we installed a Raspberry Pi on our beloved car, equipped it with a camera, designed a client, and successfully established a real-time license plate detection and recognition system. How can we create a smart car system without changing the car? For some time, the author Robert Lucian Chiriac has been thinking about how to give the car the ability to detect and recognize objects. This idea is very interesting because we have seen the capabilities of Tesla, although we cannot buy a Tesla immediately (it must be mentioned that the Model 3 is becoming increasingly attractive), but it has…

……Long press the QR code to access the original text

Machine vision innovation promotes automation forward-G 2022-03-22

Robots and advanced automation technologies such as machine vision have helped companies in multiple industries improve productivity, enhance efficiency, and increase revenue while liberating humans from repetitive and tedious tasks to increase the value of different areas in factory workshops. Although so-called blind robots can perform certain pre-programmed tasks, adding vision provides a whole new level of flexibility and opens the door to new and valuable automation tasks.

Matrox Imaging’s deep learning classification tool can detect holes, dents, and foreign objects in wire mesh. It classifies each filter as