What is an Image Feature Vector?

The feature vector for each image is an array of 4096 floating point numbers (floats). Each of those floats represents a feature of the image, such as color attributes, sharpness, or curviness. However, many of these features are far more abstract and would not be qualities a human would use, at least consciously, to distinguish images. For almost any use of these image features, it shouldn't matter which features applies to what characteristic -- you'll just use the entire feature vector and let the algorithm determine which features are most important for your task.

Our Image Features API uses convolutional neural networks to create feature vectors. The numbers in the vector represent abstract observations about the image that aren't necessarily how us humans would think about an image. These feature vectors are used to classify or compare images. For example, the cosine similarity of two images is a very good proxy for how similar the two images are. This distance measurement captures similar colors, textures, and shape.


Did this help answer your question?

thumbs up
thumbs down

Thanks for the feedback! 🙏🏽


Help by drift