Visual Machine Learning
New to AI and machine learning? Healgoo ML now offers an easy-to-master user interface and a simplified workflow for begginers.
The Healgoo ML platform allows you to build up your own AI algorithms without having to worry about the technical details.
No computer coding skills are needed...
1. Sign Up
Sign up and start your journey of machine learning with Healgoo ML.
Upload your own labeled images to the trusted HIPAA-Compliant cloud platform.
Use your labeled images to teach Healgoo ML the concepts you care about.
The simple user interface quickly tags images with your new deep learning model.
Deep learning is part of a broader family of machine learning methods based on data representations. Currently, Healgoo ML aims to solve image classification problems using deep learning technology and Healgoo's latest experiences of supervised learning and transfer learning.
Take a look at the animation on the right: if you already have a collection of images in different categories (e.g. cats and dogs), and need an automated program to predict the category for new images based on the knowledge of existing images, Healgoo ML will help.
At present, Healgoo ML is only capable of Image Classification. If you would like to stay informed about future problem solving capabilities, subscribe to our newsletter.
Most deep learning frameworks require you to prepare a set of image files and a corresponding spreadsheet file for all image-label mappings. Healgoo ML provides a more convenient way by allowing you to place the images in different label folders.
Healgoo ML also provides an executable utility* to help you upload organized images from your desktop, rather than from a web browser.
Healgoo ML also hosts public datasets on the cloud. You can test them before using your own data.Note *: available for Windows, Mac OS and Ubuntu Linux
Once the images are uploaded to the cloud, you can select a deep learning network to train your model. Rather than designing a network from scratch, you could benefit from the well-known and proven deep learning networks that Healgoo ML utilises and use transfer learning to speed up your training.
The following networks are / will be supported on the Healgoo ML platform. If you are not sure which network to choose, just use the default one.
|DenseNet121 **||8,062,504||121||33 MB||0.918|
|DenseNet201 **||20,242,984||201||80 MB||0.933|
|ResNet50 **||25,636,712||168||99 MB||0.953|
Note **: will be supported in future releases
Each network contains some mandatory fields that you need to configure (aka, hyper parameters) before training. For example, data augmentation, learning rate, optimizer and loss function etc.
But don't panic. Healgoo ML has already set default values for you based on best practice. You just need to click on the Next button and let the training begin.
Healgoo ML offers a cluster of shared GPU (Graphics Processing Unit) instances. All users' training requests will be queued and processed under "first-come, first-served" policy. During training, the model with best validation accuracy will automatically be saved. As soon as your training is completed, Healgoo ML will send you a notification email.
To maximize resource usage among all users, each training will be limited to 2 hours computation time. Based on previous training experiences, this should be enough for small or medium sized datasets (50,000 images).
If you need more computation power, please contact us for dedicated high performance GPU instances. Alternatively, it may be of benefit to use our offline machine learning products.
One of the biggest challenges for model training is to set the appropriate stopping point. If the training stops early, the model may not achieve the highest possible performance; if late, the model may enter the over-fitting stage.
The traditional way is to continuously watch the performance curve and stop the training at the correct time point (epoch).
Healgoo ML offers a series of selected Training Stoppers which evaluate the performance in the background as training occurs. They are a smart and effective way to help you finalize the model with maximum performance.
Unlike other deep learning platforms which store your trained models online and only provide prediction API for you, Healgoo ML recommends you to download models to your computer and use them for offline prediction. This will protect your business data and make your models more flexible.
Healgoo ML currently supports deployment of your trained model to a desktop application* or a mobile app**. GPU hardware and Internet connectivity are not required for prediction.
For third-party integration or if you need to use your trained model as a prediction service in local network or on the Internet, please contact us.
Note *: available for Windows, Mac OS and Ubuntu Linux
Note **: available for Android
One of the most debated topics in deep learning is how to interpret and understand a trained model – particularly in the context of high risk industries such as healthcare.
Healgoo ML offers a built-in functionality that visualizes each prediction provided by your trained models.
|Try the Visual Interpretability feature with your own image.|
Subscribe to our newsletter
Subscribe to our newsletter to get the latest updates on new releases, tutorials, and other great content related to the Healgoo ML platform.