Import your dataset
Select the dataset that you want to work within your container.
The platform will automatically group your data into clusters based on latent structures and metadata identified by the system. The closer the two clusters are, the more closely related they are in data characteristics. Each individual cluster can also be broken down into further subcategories, if desired.
Creating Your Novelty Training Set
Select a data cluster that contains the image matching the criteria you wish to train your model with.
In this scenario, since we are looking to better train autonomous cars on what to do in construction zones, we need to identify images containing construction, such as a barricade on the side of the road.
Once you identify an image matching your desired criteria, you can select the “thumbs up” to give the model the positive reinforcement to return images closely related to the desired frame.
In tandem, you may also select “thumbs down” on non-relevant images to tell the model that you do not want images matching these criteria (such as an image with no visible construction zone or images taken at night).
Reviewing and Refining Search Results
The model will begin providing you with more images that it deems appropriate to your search criteria.
At no point in this process has it been expressly defined to the model that we are only wanting images with a construction zone or barrier, but it will begin identifying these images based on the metadata.
By continuing this process of providing positive and negative reinforcement, the model will continually advance and begin identifying more accurate results. Further, by providing better data from the beginning of the training process, you can greatly accelerate the time needed to achieve model accuracy and reduce the number of iterations needed.
Saving the Results
So next time you want to look at another dataset for this example, you can get it with a push of a button!