CrowdAI named 2022 Gartner Cool Vendor in AI for Computer Vision
CrowdAI has been named one of Gartner’s Cool Vendors in AI for Computer Vision (CV) for innovative and disruptive companies.
Training a new computer vision model from scratch is no small feat. Now that you have one, it's time to put it to work labeling images for you—after all, that's the point of training a model in the first place!
Sometimes, you just want to run a model on some media it hasn't seen before. Maybe you're just testing the model out, or you just need some model results quickly and don't want to bother with GPUs, an API, containers, or a more complex set-up. Well, we have a solution for you!
Introducing CrowdAI Batch Prediction
Batch Prediction is a new feature in the CrowdAI platform that lets you take a model you trained and have it process (or "predict" on) one or more images/videos already in the platform. CrowdAI will automatically handle all of the behind-the-scenes work in the cloud so you don't have to: no need to worry about GPU set-up or figure out how to get the right media fed into the model at the right time.
With Batch Prediction, you can have your trained models process media directly within the platform—one at a time, or entire Datasets all at once. All of your results will be stored together so they're easy to come back to at any time.
Every "batch" of media you process with the model is stored together in your Project, so you can always come back to visualize the results directly in the platform. Our visualization tool allows you to overlay different types of model outputs directly on the image or video they belong to. You can view the "raw" model output as a single layer, or view the probability heatmap to get a more nuanced understanding of model performance.
Of course, you can also export the model results and take them with you!
Once your model is done processing the new media, visualize the results directly within the CrowdAI platform, so you can check quality right on the spot.
Batch Prediction is most useful when time is not of the essence. If you need real-time or near-real-time results from a model, you'll want to use our API or container prediction solutions. We'll have more on those in an upcoming post.