Do you have a quality control problem?Would you like to create deep learning models without writing code?Are you a vision system integrator and looking for a simple but high-performance solution?
AI-go
No-code Artificial Intelligence platform for manufacturing
→ A wide range of pre-trained models.→ High performance even with just a few examples.→ Just a few minutes to go into production.→ Usable even without specific skills.
Try AI-go now!
Use AI-go for 7 days and create your own deep learning models in a few minutes and with just a few examples.
AI-go provides a wide range of pre-trained models for industrial inspection and quality control: classification (binary, multiclass, and patch), anomaly detection, segmentation, and OCR (optical character recognition).
Thanks to the ability to use unsupervised models (that are trained only with examples of good parts, with no need for scrap examples), AI-go drastically reduces the time for implementing the solution (from weeks to a few minutes).
With AI-go you can achieve better performance compared to solutions developed with traditional algorithms, greater flexibility, and robustness.
Take your Vision Systems to the next level with AI-go
Join the Machine Vision GeeX Program!
An exclusive plan designed for companies with experience in machine vision, to facilitate access to AI and integrate its potential in an effective and innovative way in complete synergy with traditional solutions.
As a Machine Vision GeeX, you’ll have direct access to our special on boarding package and full partnership benefits, like training sessions with our Data Scientists and Vision Experts.
ONBOARDING PACKAGE
- AI-go Studio Full (1 year)
- Software development kit for easy integration with C++, C#, …
- Training session with our developers about AI-go Runtime integration into your application
- Training sessions with our Data Scientists and Vision Experts to highlight the power and best practices of AI
PARTNERSHIP BENEFITS
- Priority support channel (direct access to developer as well as data scientists)
- Provision of periodic training courses
- Enter the Orobix partner network opens the chance of routing commercial opportunity from Orobix to partner
- Shared co-marketing activities (digital and phisical)
Able to solve complex industrial problems
CLASSIFICATION
APPLICATIONS
Sorting defective or anomalous parts, verifying the presence of components and correct assembly, object detection.
TASK
Predicting the class of an image with confidence score.
AI-go supports two types of models:
- whole image classification: binary (good / scrap) and multi-class (defect type 1 / 2 / n)
- patch-based classification (dividing the image into smaller parts).
ANOMALY DETECTION
APPLICATIONS
Identification and localization of defects or anomalies.
TASK
Precise detection of a defect and generation of a heatmap that allows the machine operator to visualize the “anomalous” area identified by the algorithm, making the neural network’s judgment “explainable”.
AI-go learns to detect defects solely based on images of good parts (unsupervised learning), without the need for examples of scrap parts.
SEGMENTATION
APPLICATIONS
Sorting and differentiation of defects or components and products, analysis of geometries, defect measurement.
TASK
Accurate localization of a defect and generation of a pixel-level mask defining its area, accessible to the line operator for further investigation and validation of the model’s output.
AI-go supports multi-class segmentation models that allow the differentiation of defects based on categories predefined by the operator.
OCR - optical character recognition
APPLICATIONS
Reading labels and codes, especially when a highly accurate performance is required.
TASK
Recognition of printed characters even on irregular surfaces (such as vials, bottles, jars, bags, blisters, and tubes), with non-uniform print quality or embossed and dotted industrial characters.
Deep learning for visual inspection accessible to all
How it works
01. Collect images
directly from the production line.
03. Train the model on your own
just press a button!
05. Install
the trained and validated model.
START
FINISH
02. Label the images
adding the corresponding annotations to the different categories of defects (type 1, 2, n).
04. Test and validate
the model, verifying its robustness on a new data set of images.
01. Collect images
directly from the production line.
02. Label the images
adding the corresponding annotations to the different categories of defects (type 1, 2, n).
03. Train the model on your own
just press a button!
04. Test and validate
the model, verifying its robustness on a new data set of images.
05. Install
the trained and validated model.
Benefits
Automatically select the most effective pre-trained model to solve your problem and specialise it through a few examples.
Using unsupervised models – which learn from examples of good parts and do not require scrap parts – drastically reduces the time to deploy the solution from weeks to minutes!
Maintaining performance over time while ensuring an adequate level of model “explainability”.
No specific skills are required and it can be used by people without previous knowledge of computer vision, programming or artificial intelligence techniques.
Easily integrated with existing vision systems.
Models are automatically optimised to operate even on devices with reduced computational power (IoT, edge computing).
Manage the validation of new models before deploying them into production, monitor them remotely to ensure they are working properly, and schedule automatic re-training to improve their performance over time.
Designed to operate in industrial scenarios, considering aspects usually ignored by generalist solutions such as cycle time, inference speed and reliability.
Case studies
CRACK DETECTION, AUTOMOTIVE
Artificial intelligence for pinion quality control. Identification and localisation of cracks.
ANOMALY DETECTION, MANUFACTURING
Detection and localisation of anomalies on the surface of electronic boards.
QUALITY CONTROL, PHARMA
A new generation of inspection machines, equipped with artificial intelligence for vial visual inspection.
DEFECT DETECTION AND CLASSIFICATION, AUTOMOTIVE
Automatic detection and classification of surface defects on small metal components for the automotive industry.
OCR - CRIMP PRINTING, PHARMA
OCR control to read the lot number, printed with dotted font, on the curved surface of vials necks.
QUALITY CONTROL, MANUFACTURING
Artificial Intelligence for in-line quality control of metal and rubber components.
Is your application particularly complex and you need support?
Do you require a tailor-made model for your specific use case?
Lack internal expertise to design the suitable vision system for your application?
Orobix can assist you with comprehensive consulting services, providing expertise and experience to help evaluate the most appropriate solution for your specific needs.