This technology allows you to recognize objects in images—typical faces or numbers. Image recognition is most common in cases where you use AI to make products or services better or develop new services.
One such famous case is from Stanford, where scientists use image recognition to screen for cancer. Their AI can more effectively identify cancer cases, among other things because the image recognition part of their solution can see visual patterns in cells that, in a context with other indicators, can help diagnose cancer at an early stage.
Standford, cancer and AI
At Stanford, an AI model had been trained to identify if dangerous birthmarks could develop into skin cancer.
The model’s results were tested against the best dermatologists’ diagnoses. Not only did the AI model have greater accuracy in its assessment of whether the birthmark was a skin cancer precursor or not.
AI is more precise than humans
The AI model also ended up having 12 different criteria that it used in the skin cancer assessments. The dermatologists use 9. The machine has thus found three new factors which were an indication of skin cancer, which importance the best dermatologists were not aware of.
The models they used at Stanford were not “hindered” by our human professionalism or our scientific view of how a given subject – in this case, dermatologists – perceives reality.
An objective and rational AI Model
The AI model approach was objective and rational in their understanding of the given subject (Skin cancer) that they were trained to understand.
The models related to the patterns they saw in the data that were available to them and, based on this, form their perceptions of how skin cancer versus non-skin cancer could be described. That is why they came up with 12 areas that could define skin cancer and not the nine that we, as human beings, have chosen to use.
Other Use Cases
If Stanford were a traditional commercial enterprise, then they would be able to exploit this handling of complexity to gain a tremendous competitive advantage.
They would be able to make a better product and perhaps faster than competitors to find the potential cancer patients.
Notes & sources:
If you are a tech expert and what to know how to work with Image Recognition, then we suggest you read this book: Programming Computer Vision with Python by Jan Erik Solem.