“AI image recognition” in which AI identifies and distinguishes subjects in image data. With the advent of deep learning , one of the machine learning methods , AI has made rapid progress as it has become possible to learn unstructured data.
In recent years, we will explain the mechanism of “AI image recognition”, a technology that has been attracting attention in businesses in various industries and industries, and examples of its use by industry.
What is AI image recognition?
AI image recognition is a technology that uses AI to identify and distinguish human faces and characters in images . By loading a huge amount of image data into AI and learning it repeatedly, it will be able to identify various information in the photo.
For example, in 2012 Google announced the results of a study called ” Google’s Cat “. A notable point of this research is that AI was able to distinguish cats without being taught by humans.
Google has announced that it has learned AI from 10 million images and has become able to recognize and classify features in images spontaneously. As a result, it was demonstrated that AI can recognize features without human assistance . Starting with this research, attention to deep learning, which is one of the methods of machine learning, has increased, and research has been conducted all over the world.
Mechanism of AI image recognition
I will explain the mechanism of AI image recognition step by step.
① Image processing/extraction
In order for AI to recognize images, it is first necessary to input image data into a classifier (computer program that can recognize images) and let it learn. When training the classifier, image processing and extraction are performed to make the image data easy to read by the classifier.
This process not only adjusts the brightness and color of the image and removes noise and distortion, but also extracts the outline of the subject and distinguishes it from the background.
(2) Object recognition using deep learning
AI image recognition uses a method called deep learning, which is the most representative AI learning method.
Deep learning is a machine learning technology that combines simple functions to create a highly expressive “deep function” and estimates its parameters from data.
Deep learning is a learning method that utilizes neural networks, a typical algorithm of machine learning.
Neural networks are AI modeled on neurons (nerve cells in the brain)
As long as there is enough training data, the neural network itself can automatically extract the features of the data group.
There are several types of deep learning, but the model used for image recognition is called CNN (convolutional neural network) . It has a high pattern recognition ability for images and is characterized by being able to identify quickly.
A CNN consists of a “convolution layer” that extracts image features and a “pooling layer” that analyzes the features .
In the past technology, machines identified features designed by humans. However, by using CNN, AI can learn a huge amount of features and identify subjects without humans defining features in advance.
Types of image recognition technology
Image recognition technology is applied in various fields such as face recognition and character recognition.
face recognition
Face recognition is a technology in which AI extracts and identifies the features of a human face.
Technology that uses AI to extract and identify human facial features. Not only the eyes, nose, and mouth, but also detailed features are extracted from the images and videos captured by the camera. It is used for face recognition, etc., which compares face photos registered in the database.
character recognition
Character recognition is a technology that uses AI to identify handwritten characters and characters printed on paper.
A technology that uses AI to identify handwritten characters and characters printed on paper. By decoding the scanned image data with character recognition, it is possible to convert handwritten characters and printed characters into text data.
Use cases of AI image recognition (by industry)
AI image recognition has been introduced in various fields. Here are some examples of use by industry.
Defective product inspection (manufacturing)
Kewpie has introduced AI image recognition to inspect defective products such as diced potatoes used in baby food.
Initially, we considered a method to create training data from both defective and non-defective products, but created training data for only non-defective products to cover the lack of defective data. By detecting all items that do not qualify as “defective” as “defective”, the system has improved dramatically.
IoT skin care (beauty)
Shiseido is developing “Optune” that realizes personalized skin care using machines and apps.
By simply taking a picture of the skin using the camera function of a smartphone, it is possible to know the condition of the skin, such as moisture content, texture, amount of sebum, and the conspicuousness of pores. In addition, skin condition and skin care record data is collected and accumulated and used for subsequent skin care recommendations.
Unmanned convenience store (retail)
Face recognition and image recognition of products are also used in the unmanned convenience store “Smart Convenience Store” in the robot hotel “Henn na Hotel” that Huis Ten Bosch is working on.
Just by registering your face at the store entrance and arranging the items you want to purchase on the checkout table, your image will be recognized and your details and payment amount will be displayed. After checking the details, you can purchase after receiving face authentication. Face recognition is also required when leaving the store.
Prevention of dangerous driving (Maas)
Nippon Car Solutions and NTT Communications have succeeded in automatically detecting dangerous driving using AI image recognition.
By using AI to analyze multimodal data recorded by drive recorders, etc., it is now possible to automatically detect dangerous driving that could result in a collision accident, such as a bicycle jumping out, with an accuracy of 85%.
Automatic entry of handwritten slips (transportation)
Sagawa Express utilizes image recognition in a system in which AI replaces the manual work involved in inputting delivery slips.
With this system, we succeeded in reducing 8,400 hours of manual work per month. It is now possible to make effective use of resources such as time and people.
Automatically measure the human body and recommend the appropriate clothing size (apparel)
Unimate, which develops the rental uniform business, has developed an automatic measurement application “AI x R Tailor” that utilizes AI image recognition.
In the conventional self-reporting measurement system, it is easy for sizes to be different due to human error, and the maximum return rate has been recorded as 40%. Therefore, we developed an automatic measurement app that uses AI image recognition to create a mechanism that allows you to grasp the exact size before paying for the uniform. Contributes to cost reduction and service improvement.
Efficient capture and identification of data in receipts (marketing)
Cashbee Data has developed Japan’s first cashback service “CASHb” app that collects purchase data in receipts.
We acquire and use consumer purchase data from receipt images sent by users, and provide opportunities for B2C companies to utilize the data.
Fetal cardiac ultrasound screening using AI (medical)
RIKEN has developed a system that automatically detects fetal heart abnormalities in real time using AI image recognition.
This system utilizes an AI technology called “object detection technology” that can determine the position and classification of multiple objects in images with high performance by learning from training data, even in rough ultrasound images. It is useful for the treatment of congenital heart disease, for which early diagnosis during the fetal period is considered important.
Summary: AI image recognition is being used to improve operational efficiency and develop new businesses
AI image recognition has been introduced in various industries and industries. In addition to contributing to cost reductions in terms of human and time resources and improving operational efficiency , image data collected and analyzed by AI can be used to create new value and be used in new businesses and services. I’m here.
