Apple’s Machine Learning Researchers Have Developed A No-Code AI Platform Called ‘Trinity’ For Complex Spatial Datasets

17502
Source: https://arxiv.org/pdf/2106.11756.pdf

Apple’s machine learning research team has developed a no-code Artificial Intelligence (AI) platform called Trinity. This AI is designed to enable machine learning researchers and non-technical geospatial domain experts alike to experiment with different signals or datasets in order to solve problems on their own, such as complex issues that arise from the world around us every day. The ability to solve diverse problems is made possible by transforming complex Spatio-temporal datasets so that they can be consumed and solved with a standard deep learning model, like Convolutional Neural Networks. This new way of looking at data has the potential to formulate disparate problems in one standardized form for easy consumption.

Trinity is a powerful software platform for domain experts to share the stage with scientists and engineers in solving business-critical problems. Trinity has an intuitive user interface, feature store that hosts derivatives of complex feature engineering, deep learning kernel, scalable data processing mechanism.

Trinity is a tool that can be used to make AI tools accessible for everyone. Trinity lowers the barriers of accessibility by standardizing model building and deployment, quickly prototyping in order to experiment with rapid experimentation with time-to-production lowered due its efficiency as well as showcasing sample applications which will motivate others into using this new technology.

Trinity tackles complicated data problems by bringing together disparate datasets, standardizing the process of solving these challenges and providing a code-free environment to lower barriers for entry.

https://arxiv.org/pdf/2106.11756.pdf

The deep learning kernel is at the heart of this platform and encapsulates neural net architectures for semantic segmentation, providing models to keep up with ever changing data. The current implementation in TensorFlow can easily be swapped out for other frameworks as needed depending on user preference or availability.

Trinity is a versatile detection algorithm that has been used for many types of applications. Some examples include: driving behavior based road center-line detection, stop signs detected from heading profiles, type of roads seen with temporal embeddings and satellite imagery or other modalities like the car’s footprint on the ground etc.

Benefits for the Domain Expert

  • Trinity can help domain experts to run their own experiments, something they have never been able to do before. This has the potential for great things as it will encourage more participation from these knowledgeable people in projects which need this input.
  • Trinity is a zero coding platform that allows anyone to train and operate machine learning models without any coding knowledge. Users can build, iterate, deploy their model with just a few clicks.
  • Trinity is a platform for executing experiments and analyzing data. It enables quick, rapid experimentation through its simple UI backed by powerful GPUs and Hadoop clusters.

Paper: https://arxiv.org/pdf/2106.11756.pdf

Source: https://machinelearning.apple.com/research/complex-spatial-datasets