Medical AI Foundations Checkpoints: Additional Terms of Service
To use the Medical AI Foundations Checkpoints (the “Checkpoints”), you must accept (1) the Google Terms of Service, and (2) these Medical AI Foundations Checkpoints - Additional Terms of Service (the “Additional Terms”).
Please read each of these documents carefully. Together, these documents are known as the “Terms”. They establish what you can expect from us as you use our services, and what we expect from you.
The Checkpoints are intended to be used in research applications to support more data-efficient onward fine-tuning of ResNet classifications models, which you may train to perform classification tasks in chest X-ray images or digital pathology slides. Based on our research, fewer examples may be needed in your model training to achieve the same level of classification accuracy if you use the Checkpoints than if not.
The Checkpoints are for research use only.
Do not use the Checkpoints to provide patient care. For example, you may not use the Checkpoints as or in a medical device.
Do not use the Checkpoints if your use of the Checkpoints, including your sharing of content with the Checkpoints, would conflict with any commitment you’ve made to a third party.
Do not use the Checkpoints to recreate any training data used to develop the underlying model.
If you have questions about uses of the Checkpoints, contact us at: email@example.com.
Suspension or Termination
We may suspend or terminate your access to the Checkpoints if we reasonably suspect your use may cause harm to a third party, such as a patient.
If you make any publication related to your use of the Checkpoints, please cite our research behind the Checkpoints. However, you may not use Google’s name to endorse or promote your work without our permission. To request our permission, contact us at firstname.lastname@example.org.
In addition to the warranty disclaimer in the Google Terms of Service, we want to make you specifically aware of the following limitations of the Checkpoints:
- Google does not guarantee that a machine learning system initialized from these Checkpoints will perform well on any specific task or with any particular dataset (including the datasets used in the development of the Checkpoints, which are identified in this README file, but may be changed from time to time without notice to you), or recommend this initialization for any specific use case. It is your sole responsibility to build and test the final machine learning system and to evaluate performance, accuracy, reproducibility, and safety.
- Google does not guarantee that the Checkpoints have been evaluated in prospective clinical studies.
Back to Medical AI Research Foundations: A repository of medical foundation models v1.0.0