Computer Vision

Computer vision technology uses algorithms to process, analyze and extract meaningful information from digital images in order to allow computers to understand images much like humans do. It makes it possible to use automate processes that were once done by humans, opening up incredible opportunities for a wide range of applications in various industries.

Sciridge applies the utmost frontier knowledge and techniques in computer vision to automate workflow and develop smart solutions. We help businesses optimize operational processes to cut costs, and we free professionals from repetitive work to focus more on high value-added tasks.

Computer vision applies various methods in image processing and artificial intelligence to understand and predict visual inputs.

Image processing and visualization

Digital images are usually processed using a wide range of algorithms to perform tasks like image enhancement, segmentation, registration, and multi-image stitching to prepare them for further quantitative analysis. Statistical analysis and feature extraction that require complex computations are then performed to extract higher-level information. It has been applied in many fields including medicine, security, remote sensing, etc.

The quantitative data from image analysis and computational simulations are displayed through image visualization. Image visualization allows the display and interactive manipulation of digital data to represent meaningful information.

For example, in the application of rock image analysis, the workflow includes image processing, feature extraction and fluid simulation visualization.

image processing

Image Processing

image analyses

Feature Extraction

image visualization

Simulation visualization

Deep learning in feature extraction

Within just the past few years, artificial intelligence, particularly machine learning, has become far more effective and widely available. Deep learning algorithms have a significant advantage over earlier generations of machine learning algorithms. The most common deep learning architecture used in image classification is Convolutional Neural Networks (CNN). It is composed of several convolutional layers alternately connected with a number of pooling layers, and can effectively and automatically extract the image features with a high classification accuracy. CNN has pushed the boundaries of computer vision. These state-of-art algorithms can be used to detect and recognize features, identify objects, classify human actions, recognize scenery, etc. It has been applied to many fields, such as medical diagnosis, information security, and facial recognition.

Basically, a large set of images (the “training set”) are labeled to teach the computer about the features to be recognized. The training sets are then fed to train the computer to generate models for a specific computer vision problem using deep learning methods. The model will the be used to recognize features in future image datasets.

The main reason that deep learning outperforms other algorithms is that the performance of the trained model will continue to improve as more data is added to the training set. This advantage allows deep learning models to achieve close-to-human accuracy in various scenarios.

Cloud Computing

Cloud computing is one of the latest information technologies, which delivers on-demand computing services over the internet, such as servers, storage, analytics, and so on. Cloud computing offers a “pay-as-you-go” plan for everything from applications to data centers. Though cloud computing services are quite new it has been embraced by a variety of companies, from tiny start-ups to global corporations, due to its reliability and flexibility.

Cloud computing has many attractive advantages for both businesses and individual users alike. Firstly, it enables employees and professionals to work and communicate remotely. They are able to store files on remote servers and share their workplace over a cloud network  instantly and then access all the data via the internet. Secondly, it can reduce cost since it offers pay-as-you-go plans; that is users need only pay for the computing resources and workloads they use. It will save the investment for local infrastructure and maintenance. Lastly, it has powerful and reliable hosting services which greatly reduces the possibility of a “traffic jam” when there is a large amount of data needed to be processed during peak usage times.