Our Technology
Our proprietary Vision framework, simplifies the process of creating computer vision solutions that scale and integrate with your existing processes. Domain experts can benefit from our partnerships with leading technology providers, including IBM PowerAI Vision, IBM Watson Image Detection, Google Vision and a host of open source providers to quickly create capabilities that suit your needs. Using a variety of tools and reports, easily see how your solution is performing and allow you mechanisms for continuous improvement.
“There is a growing and unprecedented amount of visual content created each and every day – from social media, through manufacturing to even the satellites orbiting above us. Contained inside each of those images is insight that until now has been left untapped. Computer vision presents huge opportunities to optimise business functions and drive growth.”
Our Delivery Process
Validation
Not all problems are appropriate for computer vision. We work with you to validate the approach and ensure that the technology will deliver the results required. This may involve a data hack stage to inform viability as fast as possible.
Discovery
Following a validation of the problem and the technical solution, we will discover the scale of the challenge in order to assess the technical requirements. This may include the frequency of classification, types of classification, response times required and retraining elements.
Training
The next stage is to train the model using the data collected within the Vision Framework. Part of this process may involve adapting the training images to ensure that the classifiers are trained on the right features within the images for optimal results.
Model Creation
Best practices from industry experts have been captured into the Vision Framework so that you can easily create models with very minimal technical experience. Our tooling automatically chooses the best parameters to enable the best possible outcome.
Testing
Using an automated process we will then test the output from the training and report on its accuracy and recall for the classifications trained on.
Integration & Improvement
Once developed, the resulting classifier can be integrated into existing processes using a workflow automation tool based on Node-RED and integrated into the Vision Framework. With the API integrated you can continually collect information to support new models, training sets and performance metrics. Now deployed into production, use the suite of reports to continually monitor the results in order to maximise the accuracy and respond to changes in real-world use.