Caltech-10
A 10-class image classification benchmark dataset derived from Caltech-101. It is commonly used to evaluate quantum and optical image classifiers, providing a reduced set of categories that keeps circuit depth and optical hardware requirements manageable while retaining natural-image complexity.
Timeline (last 90 days)
In simulation, four phase plates at a single wavelength achieve 72.3% accuracy on Caltech-10 color photographs, 96.8% on MNIST, 86.7% on Fashion-MNIST, and 47.9% on color CIFAR-10.
On IBM's ibm_kingston quantum computer, classification accuracy on 300 Caltech-10 test images is 37.7% versus 10% for random guessing; the same configuration reaches 56.3% in exact simulation.
All Facts (2)
In simulation, four phase plates at a single wavelength achieve 72.3% accuracy on Caltech-10 color photographs, 96.8% on MNIST, 86.7% on Fashion-MNIST, and 47.9% on color CIFAR-10.
50%UnverifiedOn IBM's ibm_kingston quantum computer, classification accuracy on 300 Caltech-10 test images is 37.7% versus 10% for random guessing; the same configuration reaches 56.3% in exact simulation.
50%