Product development /

Wizards, inspections, APIs and more new features

June 24 2020/5 min read
  • Reynaldo Boulogne
    Reynaldo Boulogne

We have been quiet over the past 8 months (since our announcement that we added a pre-trained BERT model to the platform to be precise). A lot has happened since, so it’s high time to share what we have been up to.

Here's a list of our newest features, models, and tutorials.

New features

Dataset view

  • 📊 Find useful datasets with a few clicks with our new Data library. The data library contains over 30 useful datasets applicable to different deep learning problems. They are waiting for you.
  • ⌨️ Programmatically upload your data into the platform with the Data API or by using our integrations with Google BigQuery or Azure Synapse. Meaning, no more selecting and uploading files manually on the platform.
  • 🌆A more flexible way to upload image data. Most image datasets are structured in a way such that the images are put into folders, where the name of the folder is the actual categories. This structure is now supported on the platform.
  • 🧰 Image preprocessing features allow you to resize, crop and pad your image datasets directly from the platform. No need to do this step in a notebook anymore.
  • 🧪 Test sets can now be created directly on the platform by automatically or manually specifying how your dataset should be subdivided into train, validation and test sets. 

Modelling view

  • 🔮 Get help with solving your specific AI problem with our new Experiment creation wizard. After saving your dataset, the wizard offers the most suitable neural network template based on the input data and the type of problem you’re trying to solve.

Evaluation view

  • 🔍 Prediction inspections are here! You can now check your model’s predictions on your validation / test set after training, right from the platform. This is a powerful tool that lets you profile your model and diagnose its performance in more detail than ever before.
  • 🛑 Early stopping will automatically stop your model from training further, if there hasn’t been any improvement in the training / validation loss after a couple of epochs. This will not only save you time, but also help you avoid burning through your GPU hours unnecessarily.

Deployment view

  • 👀 Easily test your model with our classifier apps directly from the deployment view. They can be used for single input image and text classification problems.

New models / snippets

  • 💨 Train your AI models faster with the new pretrained block, MobileNetV2. Perfect to use for image processing.
  • 🎯 Achieve higher accuracy and efficiency on image classification tasks with the new pre-trained state-of-the-art model, EfficientNet. EfficientNets are a family of neural networks released by Google in 2019, maximizing the accuracy for a given computational cost.

New tutorials

  • 🤖 Build your own AI bot that analyzes and classifies Slack messages from different channels and let your co-workers know when they’ve posted on the wrong channel.
  • 😃The first time I tried coding, I built an AI model. - Follow our non-data-scientist colleague on her journey of building a model that can tell the difference between Twitter users by identifying patterns in previous tweets from that account.
  • 📚Classify books by genre solely on its summary! Learn how you can use BERT on the Peltarion Platform to build a model on your own and correct all major bookstores in your country!

That was quite a list. Kudos if you made it all the way down here :-)

If you found all of this interesting, why not give the platform a try? We have a free tier so you don’t need to commit to anything beforehand. 

P.S. did you know that the platform also offers you all the training infrastructure (i.e. GPUs) and cloud deployment infrastructure that you need when you create an account? There is nothing to set up or configure, so you can get started right away. Beats even the best of laptops, don’t you think? ;)

  • Reynaldo Boulogne

    Reynaldo Boulogne

    With over 15 years of experience, Reynaldo has worked within the intersection of business and technology across multiple sectors, most recently at Klarna and Spotify. He is passionate about innovation, leadership, and building things from scratch. Reynaldo is also a former Vice-chairman of the Stockholm based AI forum, Stockholm AI.