Deep speckle correlation: a deep learning approach toward scalable imaging through scattering media

Data for training:

Letter speckle: each diffuser we use 300 images for training [image and corresponding ground truth index range:0-299)]

Digit speckle: each diffuser we use 300 images for training [image and corresponding ground truth index range:0-299)]

Data for testing:

Seen digits and letters: each unseen diffuser 400 images.

Including:

200 letters [image index 0-199)] corresponding letter ground truth index 0-199

200 digits [image index 200-399] corresponding digits ground truth index 0-199

Unseen digits and letters

Unseen digits:  100 images are used for testing [image and corresponding ground truth index range 300-399]

Unseen digits: 100 images are used for testing [image and corresponding ground truth index range 300-399]

Quickdraw : different object types