Deep Learning Add-on

Deep Learning Add-on is a breakthrough technology for machine vision. It is a set of five ready-made tools which are trained with 20-50 sample images, and which then detect objects, defects or features automatically. Internally it uses large neural networks designed and optimized by our research team for use in industrial vision systems.

Together with Aurora Vision Studio you are getting a complete solution for training and deploying modern machine vision applications.

Why deep learning from Aurora Vision?

1. Get a complete graphical environment for model training and application design.
2. There is no programming – load your images, add labels and click "Train".
3. It is optimized with WEAVER inference engine, running very fast on both GPU and CPU.
4. BONUS: You can also prepare training data with Zillin online tool.

Key Facts

Trainig Data

Learns from few samples

Typical applications require between 20 and 50 images for training. The more the better, but our software internally learns key characteristics from a limited training set and then generates thousands of new artificial samples for effective training.

Hardware Requirements

Works on GPU and CPU

A modern GPU is required for effective training. At production, you can use either GPU or CPU. GPU will typically be 3-10 times faster (with the exception of Object Classification which is equally fast on CPU).

Speed

The highest performance

Typical training time on a GPU is 5-15 minutes. Inference time varies depending on the tool and hardware between 5 and 100 ms per image. The highest performance is guaranteed by WEAVER, an industrial inference engine.

Training Procedure

Collect and normalize images

Acquire between 20 and 50 images (the more the better), both Good and Bad, representing all possible object variations; save them to disk.Make sure that the object scale, orientation and lighting are as consistent as possible

Training

• Open Aurora Vision Studio and add one of the Deep Learning Add-on tools.
• Open an editor associated with the tool and load your training images there.
• Label your images or add markings using drawing tools (you can also import data from Zillin).
• Click “Train”

Training and Validation Sets

In deep learning, as in all fields of machine learning, it is very important to follow correct methodology. The most important rule is to separate the Training set from the Validation set. The Training set is a set of samples used for creating a model. We cannot use it to measure the model’s performance, as this often generates results that are overoptimistic. Thus, we use separate data – the Validation set – to evaluate the model. Our Deep Learning Add-on automatically creates both sets from the samples provided by the user.

Execute

Run the program and see the results. Go to 1 or 2 until results are fully satisfactory.

How is it different from TensorFlow or PyTorch?

TensorFlow and PyTorch are low-level frameworks for programmers and for passionates. If you have your R&D team, you may want to use these tools to build a solution from scratch. It may take a couple of days to create a demo, but at least several months to have a production-ready system, and even more to achieve the highest possible performance. On the other hand, our product is a complete solution, field-tested in over 100 projects and you can use it yourself today.

Why should I use Aurora Vision if I could use an open source neural network?

We provide you a complete solution – it consists of five optimized neural network designs, but also of: graphical tools for easy data annotation and training, advanced augmentations and automatic balancing of training data, mix of traditional and machine learning methods for data preprocessing, optimized memory management, industrial-grade inference engine for CPU and GPU, tools for deployment, technical support + know-how. 

We spent many man-years in developing, testing and fine-tuning all of that so that you can bring it to your project instantly, with reasonably small training set, with performance much above the open-source frameworks and at a low cost at the same time.

WEAVER

WEAVER is a high performance inference engine for machine vision. It executes your deep neural networks on both nVidia GPU and Intel CPU with the highest performance. Being a fully commercial product, it assures industrial grade quality and long-term support.

Why WEAVER?

• 3x Faster 
• No Python code 
• Long-term support 
• Ease of use

Want to know technical details?

WEAVER is a library coming with both C and modern C++ interfaces. You need to convert your data structure to WEAVER's tensor (multi-dimensional array) and then you can invoke its two functions: deploy and run.

Features

Ease of use

Most of the other inference engines require you to do the Python programming and tweak many things. WEAVER is different. He only does two things: (1) model optimization, (2) execution. All you need to deliver is your H5 network file (Keras output). Moreover, when you create your solution with Aurora Vision Studio, WEAVER is already employed in its ready-made deep learning tools.

Long-term support

The fast pace of development in the field of deep learning is not particularly helpful in developing real-world systems. Most of our customers require long-term support and stable availability of the components they use. WEAVER solves that. He is independent of the open-source frameworks, eliminates Python code and provides long-term support for commercial projects.

Bespoke service

We are specialists in porting results of machine learning research to production and to other real-life environments. Our team will help in optimizing your neural networks for the given hardware and integrating it with applications written in C, C++ or C#.

Performance

Typically, WEAVER provides 3-10 times better execution time than an open-source framework without optimizations. Here you can see our internal results that we achieved for one of our ready-made tools, showing competitive performance also against other available inference engines.