EasyDeepLearning

• Neural Networks are computing systems inspired by the biological neural networks that constitute the human brain and Convolutional Neural Networks (CNN) are a class of deep, feed-forward artificial neural networks, most commonly applied to analyzing images. Deep Learning uses CNNs to solve complex problems difficult or impossible to solve with so-called conventional computer vision algorithms. EasyDeepLearning learns by example. It learns how to distinguish defects or classify parts by being shown many images of the parts to be inspected. This is called the training process.
• EasyDeepLearning does not require the user to explain how to differentiate good parts from bad ones or how to recognize products from each class; it only requires the user to label training images, that is to tell which ones are good and which ones are bad, or which ones belong to which class. After this learning/training process, the EasyDeepLearning library is able to classify images. For any given image, EasyDeepLearning returns a list of probabilities, showing the likelihood that the image belongs to each of the classes it has been taught. For example, if the process requires setting apart bad parts from good ones, EasyDeepLearning returns whether each part is good or bad, and with what probability.

Deep Learning is generally not suitable for applications requiring precise measurement or gauging. It is also not recommended when some types of errors (such as false negative) are completely unacceptable. EasyDeepLearning performs better than traditional machine vision when the defects are difficult to specify explicitly, for example, when the classification depends on complex shapes and textures at various scales and positions. Besides, the "learn by example" paradigm of Deep Learning can also reduce the development time of a computer vision process.

Deep Learning works by training a neural network, teaching it how to classify a set of reference images. The performance of the process highly depends on how representative and extensive the set of reference images is. EasyDeepLearning implements “data augmentation”, which creates additional reference images by modifying (for example by shifting, rotating, scaling) existing reference images within programmable limits. This allows EasyDeepLearning to work with as few as one hundred training images per class.

Open source versions of neural networks are available for free, so why choose Open eVision’s EasyDeepLearning?

• EasyDeepLearning has been tailored, parametrized and optimized for analyzing images, particularly for machine vision applications.
• EasyDeepLearning has a simple API and the user can benefit from the power of deep learning with only a few lines of code.

Open eVision also includes the EasyDeepLearning Studio application. This application assists the user during the learning and testing phases.

Deep Learning generally requires significant amounts of processing power, especially during the learning phase. EasyDeepLearning supports standard CPUs and automatically detects Nvidia CUDA-compatible GPUs in the PC. Using a single GPU typically accelerates the learning and the processing phases by a factor of 100.