At TrainingData.io we understand the urgent need for diagnostic tools to identify COVID-19. In the initial stages of this outbreak all countries, including the US, have faced one major problem - lack of diagnostic tools and proper testing. To be able to build diagnostic tools, data scientists are scrambling to get their hands on whatever scarse data is available.

Given the need for diagnostic tools to combat COVID-19 pandemic, TrainingData.io is providing a collaborative workspace preloaded with open-source dataset. This collaborative workspace allows data-scientists and radiologists to share annotations of training data to be used for training machine learning models for COVID-19. This dataset is seeded from GitHub covid-chestxray-dataset.

Open-source dataset for research: We are inviting hospitals, clinics, researchers, radiologists to upload more de-identified imaging data especially CT scans. The purpose is to make available diverse set of data from most affected places, like South Korea, Singapore, Italy, France, Spain, USA. This contribution to open-source community will help independent researchers to build diagnostic tools faster.

COVID-19 Dataset Annotation & Collaboration

Build ML models using NVIDIA Clara train SDK: Team at TrainingData.io has expertise in building segmentation models using U-Net on NVIDIA Clara. We would like to offer our services to build open source models for the community. Here is a sample model to detect vertebral bodies made using 2D U-Net model in NVIDIA Clara train SDK.

Segmentation model to detect vertebral bodies using 2D UNet 

Contact us: For further information reach out to us: info@trainingdata.io