Mindtech Chameleon人工智慧訓練平台

Using a GIGABYTE server, the solutions for vision and AI applications can be found using synthetic data.

Combining Synthetic and Real Data for Results

The Mindtech Chameleon AI training platform reduces the cost and time for customers to develop AI solutions while improving the quality of their products. The platform has a complete workflow for AI training data development. It delivers annotated synthetic data, which targets customers' specific needs. Also, there is an easy to use selection process for management and automation of the training data sets. Mindtech’s platform has a simulator suite and AI tools. As well, the tools work with industry standard frameworks such as Tensorflow and PyTorch. After the data is processed in the simulator it is fed to the AI tools for training.

Why Mindtech Chameleon

Complete Solution

The simulator produces annotated images for the AI tools to train, then kick out results. The platform makes the process fluid.
Simple Editing

The Scenario Editor is a simple UI based, drag'n'drop tool that creates the exact scenarios and corner cases needed in an intuitive way.
Readiness

Mindtech can supply complete data packs, ready for training AI systems, such as sim behavior, actors, and sensors.
Adaptable

Mindtech can introduce new data for changes in the environment, such as people wearing facial masks or automated vehicles on roads.
Streamlined Process

Mindtech's synthetic data has automatic generation of advanced annotations that signifcantly reduce time and cost compared to real data.
Accurate Results

The pixel perfect annotations and ability to address bias in datasets improves results accuracy.

Mindtech Chameleon AI Training Platform

Using a GIGABYTE server, the solutions for vision and AI applications can be found using synthetic data.

Retail Store

To maintain a safe distance between people in a store. Using synthetic data to train a neural network, alerts would sound if people are not keeping a safe distance in a store or queue.
Shopping Malls & Hospitals

To maintain a close, but not too close, distance between a robot and a person. A robot that is AI trained (with image classification) could be employed to clean floors or seamlessly interact with customers, through accuracy of the training.
Smart Enforcement

To find out how frequent or identify riders that do not wear a helmet. By training an AI system, it can automatically generate reports identifying how well people are following motorcycle helmet laws. It also can identify license plates of offenders and report accordingly.

You May Be Curious

What is Mindtech Chameleon AI training platform?

The platform targets AI, machine learning, deep learning, and neural networks. Through neural network training the data sets can analyze and produce results. For example, a virtual world is created using 3D graphics. (The virtual model mimics behaviors seen in the real world.) Cameras are placed in the virtual world to record the activities. The output of this is saved, along with annotations, that are then fed into the ML framework to train the AI network. The Mindtech Chameleon AI training platform can be found in GIGABYTE servers. Check with GIGABYTE to find out how.

What components/performance does my system need?

That is heavily dependent on the complexity of the scenario, the number of cameras that are modeled, the required annotations, and the amount of data required. Most servers will offer reasonable performance, if not exceed your expectations.

What are the licensed deliverables?

Customers may elect to license complete "ready-to-go" data sets which offer data suitable for training networks for specific tasks. Alternatively, the full tool suite may be licensed (along with relevant application and asset packs) for customers to create their own data.

What different scenes and scenarios can be produced?

Mindtech has a number of pre-made application packs covering typical smart city and smart vision use cases. New packs are created on demand to meet precise customer requirements.

What annotations are available?

Annotations include 2D bounding box, 3D bounding box, semantic segmentation, and range data.

What AI frameworks are supported?

The data produced is independent of framework. Conversions can be made to any framework such as PyTorch, Keras, Tensorflow, etc.

Can I use synthetic images only?

While synthetic images can be used on their own, it is typical to use a combination of real and synthetic data. The combination of the two leads to the best results in terms of accuracy. The number of synthetic images to real images will depend on the application. Synthetic data alone is useful for prototyping, and may be used for some applications in isolation.