A photo collection acquired from Unsplash dataset.
A novelty collection of 1200 horizontal images, manually annotated and equally balanced in Shallow/Deep depth of field captures. Splitted in train/validation/test sets of 560/140/500 images.
An actively selected data pool of 2000 images (1000 randomly/1000 actively) selected to simulate the active learning framework.
The collection of Unsplash lite dataset w/ 13,521 horizontal images, resized in 900x600 resolution.
The total collection of Insplash lite dataset w/ 25318 square images, from horizontal and vertical, transformed in 400x400 resolution.
Integrity checks (SHA-256):
-
unsplash-horizontal-big.zip
2e5f7ac554a0fbaed3366b785f982d0527794bbf69bafb8d3565618745382a1e -
unsplash-square.zip
cbe101ac87e0877f9e7a021b5975fde0ca4e57c712ca3d24074d3e072f2ba47d