Vision Systems for Deep Learning
Deep Learning in Image Processing
The following topics are covered in detail in our White Paper:
Fields of application of deep learning in machine vision
Advantages of Deep Learning-Based Methods
Cost of using deep learning
Optimization of deep learning networks through hybrid approach
Fields of application for solutions without deep learning
Deep learning is rapidly spreading across computer vision applications. The benefits of artificial neural networks (ANNs) are twofold. ANNs have the potential to improve the accuracy and robustness for applications in factory automation, robotics or retail. At the same time, ANNs have the capability to solve image-based application problems that could not be solved in the past, such as pathology detection in microscopy or complex pattern classification in surface detection.
Comparison of classical image processing with deep learning
Classical image processing
Possible applications:
Recognition of simple shapes and structures
Rectification and coordinate conversions
Measurements of positions, distances, sizes
Preprocessing of images
Code reading
Advantages:
Fast and easy setup
Sophisticated, traceable algorithms
Deep learning-based image processing
Possible applications:
Recognition of elements with varying shapes and sizes
Classification of complex elements and structures
Recognition with varying backgrounds
Recognition under varying light conditions
Text recognition
Advantages:
Robust setup
High performance for recognition of complex elements
FPGA frame grabber-based systems for deep learning: fastest inference and highest reliability
Best performance, fastest inference per second, highest reliability – if your application demands high throughput, the FPGA frame grabber-based vision system for deep learning is the right fit for you. The microEnable 5 marathon deepVCL from Silicon Software, together with the FPGA configuration software VisualApplets, let you deploy your ANN on FPGAs with just a few clicks!
More about FPGA frame grabber-based systems for deep learning
PC-based systems for deep learning: fast time-to-market with lowest integration costs
Lowest integration costs and fastest time-to-market: PC-based systems score with an easy design-in. Use our plug-and-play hardware and software components to build your PC-based deep learning vision system. Our broad ace camera portfolio and the pylon Camera Software Suite make it easy to deploy your ANN without spending too much integration effort.
More about PC-based systems for deep learning
Embedded vision for the most compact and cost-effective deep learning solutions
The most compact and cost-effective vision systems can be designed using embedded technology. The combination of board level cameras and embedded processing units ensures the lowest cost per unit. Intelligent edge devices deliver fast runtimes, low latency and advanced privacy and security. From camera modules to concept studies and ready-to-use solutions – let us take care of your embedded vision system for deep learning.
More about embedded vision systems for deep learningWhat types of training data are there?
While this order does not change, there are various ways in which the training data can be handled:
Supervised learning: The algorithm receives both the questions and the corresponding answers. This data is learned so that new questions only need to be assigned correctly.
Unsupervised learning: The algorithm only receives questions and learns to recognize patterns and structures from the data. By grouping the data correctly, cluster analyses can then be carried out, for example.
Reinforcement learning: The system interacts with the environment in order to learn the optimal action by acting incorrectly or correctly. Wrong actions are punished, while correctly executed actions are reinforced.
As the neural network is the core of the algorithm and contributes significantly to the success of the system, training the network is very important. It is no coincidence that large corporations such as Google, Microsoft and IBM have invested huge sums in the development of machine learning and the creation of databases. Google has even made its own TensorFlow database available as open source, which means that the database is not only publicly accessible, but is also constantly being developed further.
