Visual software

AI visual inspection software

▶ Product Introduction
Supports diverse application scenarios: Eight built-in AI functions address complex defect location, detection, classification, and character recognition. Meets core project requirements: Leveraging extensive experience across multiple industries and process stages, and continuously developing AI visual inspection algorithms, the company boasts robust detection capabilities. Zero-code model construction: Utilizing a graphical interface, the company builds a complete AI model integrating image annotation, model training, and optimization without programming. Easy to integrate: Able to be integrated into diverse production line equipment across multiple industries, facilitating quality control and improving yield.
Download product brochure

▶ Product advantages

Powerful performance and wide coverage

  • Solve various complex detection problems:

Effectively deal with low contrast, many types of defects, complex background, small defect size, etc.

Complex detection scenarios.

The minimum detectable size is 3 pixels, and can be as low as 2 pixels when the defect contrast is obvious.

  • Adapt to high production line speed requirements:

Training 100 images takes about 10 minutes, and pixel-level defect detection for a single image is possible.

Within 30ms, defect classification takes less than 1ms.

  • Support small sample scenarios:

Built-in small sample algorithm, supports good product learning mode, accurately identifies single type of defects

30-50 images are required, and even no defect samples are required.

Rich functions and high online efficiency

Helps to efficiently build high-quality models, improve online efficiency, and achieve comprehensive speed-up

65%-90%。

  • Incremental training: Only newly added defect images need to be learned, shortening the model retraining time.
  • Training set recommendation: Supports intelligent selection of effective data to improve model building effect and efficiency.
  • Factory mode: By analogy with the actual production line model detection effect and time, the causes of over-detection and missed detection can be quickly located to improve model verification efficiency.

Flexible secondary development and low overall cost

  • Provides a standardized interface, compatible with mainstream operating systems and common industrial vision software.
  • Supports multiple programming languages, allowing for flexible secondary development for special scenarios.
  • A single image can detect multiple defect types and make comprehensive judgments without the need for secondary development, reducing time and labor costs.
▶ Features
▶ Product Architecture
▶ Functional modules

Split

Perform pixel-level inspection on images to accurately identify the location, size, and type of defects.

Application scenarios: Detection of small defects on product surfaces and defects in complex scenarios

Unsupervised segmentation

Only good product images are needed to perform pixel-level inspection for all known and unknown defects, allowing for rapid online verification.

Application scenario: high production line yield, long sample collection cycle and unknown defects

Detection

Perform area-level inspection on images to accurately identify targets or defects.

Application scenarios: multi-target detection, multi-class defect detection

position

Detect the category, location, size, and orientation of single or multiple objects in an image.

Application scenario: high-precision target positioning

Character Recognition

Efficiently recognize text in images.

Application scenarios: overlapping/distorted/skewed character recognition, complex background character recognition, multiple font and text type recognition

Assembly inspection

Detect the presence or absence of targets in a fixed area and whether the quantity meets the requirements.

Application scenario: Quickly verify whether the assembly meets the set standards

Classification

Determine the category of the entire image.

Application scenarios: defect classification, product selection and grading

Unsupervised classification

Only good product images are needed to fully classify all known and unknown defects and quickly conduct online verification.

Application scenario: Anomaly category detection on the production line