SensorMODEL™
At the heart of all of our core visual sensing products is a real time Computer Vision processing pipeline supported by an ever-growing catalogue of pre-trained Deep Neural Network models.
Many of these Computer Vision models are created by our expert in-house team using skills, cutting edge tools and expertise built up over years of advanced Research & Development to solve real world problems for our clients.
Our ability to rapidly train and deploy new models into our sensing platform from highly-specific, contextual datasets is a key part of our competitive advantage.
Now we can offer this unique capability to help you deliver your projects and business processes faster and at lower cost.
What is a custom Computer Vision model?
A custom Computer Vision model is a Machine Learning model that is specially designed and trained to perform a specific task related to image or video analysis. It is tailored to your unique needs and requirements.
Custom Computer Vision models are typically built using deep learning techniques, such as Convolutional Neural Networks (CNNs). These models are trained using large amounts of labelled data, such as images or videos, to learn the features and patterns that are relevant to a specific task.
Your custom Computer Vision model will be trained in your required visual sensing task using high-quality labelled training datasets, which is images and ‘metadata’. These are typically provided by you but in cases where you have the images but not the metadata, our own data annotation and model generation team will work with your subject matter experts to gain knowledge of the image dataset in order to create the most accurate and efficient model for your use case.
After we’ve delivered and deployed your trained custom Computer Vision model into the runtime environment we can organise to incrementally retrain and redeploy it when you supply new labelled training data from time to time. This will ensure that your custom Computer Vision model will be kept up to date and accurate as your business or project evolves.
The trained model produced can be reused for repeated inferencing on image or video data supplied to it in either real time or from a batch of images supplied to it by your team.
Your trained model can be made securely accessible as a ‘Software as a Service’ fully hosted and operated on your behalf in our platform. With this option the model will be available to multiple users within the your ecosystem in a ‘self-service’ fashion with the ability to integrate into your applications using a secure web API that we provide.
