Rendered.ai offers a service for comparing the imagery between two datasets using a dimensionality reduction technique. The way our service works is that we first gather features using a Feature Pyramid Network to determine what Object Detection models are seeing in images at different levels of feature size. Then we reduce those features using UMAP as a dimensionality reduction technique for visualizing datasets in a 2D or 3D space. We generate an interactive 3D plot where users can click on data points to view images. Using this, users can compare the imagery of two or more datasets and infer what is making them similar or different.
Creating a UMAP Job
To start, we will need to navigate to the Dataset Library page of our workspace and select the datasets we would like to compare, then click the compare button.
In the next dialog we can name our UMAP comparison, determine which dataset will be used as the fit dataset and how many images to sample from each dataset for comparison.
Clicking on the Compare button will start a new UMAP job.
Just as with the other dataset services, the icons for job status has the following meanings:
No symbol means that the service job is complete and ready to use.
The sand dial symbol means that the job is running. It will remain this way until the job has either completed or failed.
The error symbol means that the job has an issue. You can click on the symbol to fetch a log of the service to help determine what caused the issue.
When the job is complete the status symbol will disappear and there will be new icons for go-to, edit and delete the UMAP job. The following image shows a completed UMAP comparison for that dataset.
There are two ways to view the UMAP comparison, just like the Analytics microservice. The first is to click the go-to icon on the UMAP section of the Datasets Library page as shown below.
The second way is to navigate to the Analyses Library and click on the same icon.
Once the UMAP page has loaded, it will show an interactive 3D plot. You can click on each data point to get information about the image.