As simple as drawing a cell
Three new output modes: masked channel, segmentation and smart segmentation
Create objects in 2D or 3D automatically with either segmentation modes
Smart segmentation extracts physical parameters from your drawing for object detection
Zero image analysis expertise required
Aivia 8 adds three additional output modes to the Pixel Classifier: masked channel, segmentation and smart segmentation.
Masked channel output creates a new image output from the original image excluding the regions below detection threshold. Segmentation output creates 2D or 3D objects based on the detection threshold.
In smart segmentation mode, Aivia interactively learns from the users' drawn examples to automatically detect and partition objects of interest. The results are previewed live as you draw on the image. With smart segmentation, there are no image analysis parameters to learn or adjust, just paint a few examples of cells of interest and let the software handle the rest.
Mesh creation on the fly
In Aivia 8, we are adding a new way to create meshes manually with the Mesh Editor. Choose any slice of the 3D volume and start painting. The mesh generates automatically and updates as you paint. Predictive auto-draw and auto-fill tools help you expedite the mesh creation process by automatically creating the annotations on the next slice you visit. You can edit the pre-computed annotations and Aivia will automatically adjust the auto-detection region on the following slice.
You can start painting the next object as soon as you complete the current one thanks to the fast, on-the-fly mesh creation algorithm. The painted slices can be exported as ground truths for the deep learning pipeline.
Real-time mesh painting and editing in 3D
Create multiple objects in parallel
Predictive auto-draw mode generates annotations in adjacent z slices based on initial user input
User edits, while in auto-draw mode, positively influence predictions on the following z slice
Auto-fill and auto-step to quickly progress through z
Export manually-created meshes as ground truths for deep learning
Tap into the cloud
Aivia 8 is designed to work seamlessly with Aivia Cloud and is the first software solution providing an end-to-end deep learning image analysis workflow. Aivia 8 provides the tools you need to efficiently create and/or edit the ground truth data required to train your own AI model using cloud computing. Aivia Cloud enables you to easily train or apply deep learning models for image enhancement, organelle prediction and image segmentation. NO coding, image analysis or machine learning expertise needed.
Aivia Cloud leverages the power of Google Cloud Platform (GCP) to bring you state-of-the-art hardware for training (and applying) deep learning models. You can train your own deep learning model in hours instead of days on Aivia Cloud without the hassle of buying and maintaining your own hardware. Applying a deep learning model takes minutes instead of hours or days. Aivia Cloud includes all the necessary Keras, TensorFlow and Python dependencies so you don't need to worry about any of it (i.e. installation, set up, compatibility).
The results and deep learning model can be downloaded to your desktop and reviewed in Aivia 8. No additional conversion steps needed.
First software to provide end-to-end deep learning image analysis
Train (including transfer learning) your own AI model
Apply one of the pre-trained models or one trained with your data
Download and review analysis results in Aivia 8
Load the deep learning model locally to apply to other datasets
The mesh creation and rendering pipelines have been optimized to further benefit from GPU computing. Meshes in Aivia 8 also includes multiple levels of detail (LOD) that enables fast loading and display of meshes - even when GPU resources are limited.
From Aivia 8 meshes include multiple levels of detail (LOD). Meshes of nearly unlimited size load instantaneously (less than 5 seconds) at the lowest level of detail followed by the highest LOD allowed by your video memory. This enables you to see the rendered meshes even when the graphic resource on your computer is limited.
The novel approach introduced with Aivia 8 - mesh LOD - enables file loading up to 2.8x faster and mesh rendering is up to 2.2x faster than Aivia 7.7.
Completely revamped mesh creation and rendering pipeline using GPU processing
Up to 3x speed improvements in mesh creation
Up to 2.2x speed improvements in mesh rendering
Multi-resolution mesh level of details (LOD)
In Aivia 8, marching cubes and connected components algorithms use GPU in place of CPU for mesh creation. GPU processing enables extra parallel processing capabilities over CPU processing that yields up to 3x speed improvements in mesh creation over Aivia 7.7.
Please fill in your contact details and tell us more about your research and image analysis requirements
15405 SE 37th St, Ste 100
Bellevue, WA 98006