A Powerful Tool for Image Analytics and Model Management
Streamline your imaging workflow with Retina’s deep learning image analytics and model management application. Retina offers a wide range of deep learning tasks throughout the imaging pipeline, including model management, dataset curation, image annotations, preprocessing pipelines, model training and inferencing, and model deployment. Built on top of the Layar data fabric architecture, users can derive insights from their images in the cloud or on-prem, making this an attractive solution for companies working with multiple imaging sites and collaborators.
Learn more about how Retina can simplify the way you develop models for deep learning image analytics within your organization.
Prepare Images for Machine Learning Tasks
Normalization & Augmentation
Retina provides a built-in augmentation library for image modification to enhance deep learning model training accuracy. Reduce artifacts and accentuate natural features in an image to improve precision further down the pipeline. Retina can also facilitate de-identification tasks, such as blurring, censoring, or redaction of sensitive information embedded in images.
Diverse Sampling and Tiling Strategies
Optimize deep learning models with tiling and Multi-Instance Learning techniques to minimize network/memory footprint and facilitate training on large images like whole slide images (WSI) for pathology.
Designated Space For Your Team of Experts to Label Images
Clinicians and domain-experts can further annotate images with tags specific to the project for a deeper understanding of the image contents. Have your team of pathologists review a series of whole-slide images to identify potentially cancerous tumor samples, and this metadata can be utilized to facilitate information retrieval or to develop deep learning models.
Deep Learning Models for Image Classification Tasks
Annotate and assign labels to input images using Retina’s image classification pipeline. There are several classification use cases, including the prediction of malignancy in tissue sample images, or the classification of crystal morphology (ie. hexagonal, granular, etc.) for pharmaceutical quality control.
Diverse Image Format Compatibility
Compatible with an life science data type portfolio.
Vyasa technology is compatible with a diverse number of image file types, including but not limited to file formats commonly used in medical imagery and microscopy. We’ve included a number of standards as well as the most common proprietary formats, to offer simplicity in the upload and processing steps of the pipeline.
Model Management Dashboard
A High Level View of Models in Testing & Deployment
Manage, analyze, and store model metadata within a dashboard, for easy deployment. Deploy your models to decentralized sites and maintain oversight of their performance with scheduled health checks and quality control metrics.
Dynamic Compute Technology
Greater Efficiency With CPU/GPU Parallelization
Retina automatically detects the amount of compute power required and can smartly switch between GPU and CPU hardware as needed. Users can train multiple models in parallel across several GPU instances, and we will work with you to create an efficient parallelization strategy for your imaging tasks.
Detection of breast cancer on screening mammography is challenging as an image classification task because cancerous tissue only represents a small portion of the tissue in the image. Retina rises to the challenge with localized tiling to deliver state of the art results.
Deep Learning Image Analytics Application
An image analytics application that offers a wide range of deep learning image-related tasks, including management, annotation and deep learning analytics on images.
Find more support in our help center about:
- The Basics of Retina
- Creating a Classification Project
- Annotating Images in Retina