new icn messageflickr-free-ic3d pan white
t-SNE visualization of animals | by Gene Kogan
Back to photostream

t-SNE visualization of animals

a gridded 2d visualization of 1,728 pictures of animals clustered by similarity using t-SNE. each of the images are first encoded into 4096-bit feature vector derived from the activations of the last fully-connected layer in a convolutional neural network. Then the features are reduced to 2d using t-SNE dimensionality reduction technique, and then gridded using RasterFairy.


software used:


images downloaded from Caltech-256:


animals include:

bat, bear, bonsai*, butterfly, cactus*, camel, centipede, chimp, cockroach, conch, crab, dog, dolphin, duck, elephant, elk, fern*, frog, giraffe, goat, goose, gorilla, grasshopper, hawksbill, hibiscus, horse, horseshoe-crab, house-fly, iguana, hummingbird, ibis, iris, killer-whale, kangaroo, llama, mushroom*, octopus, ostrich, owl, penguin, porcupine, praying-mantis, raccoon, scorpion, skunk, snake, snail, spider, starfish*, sunflower, swan, greyhound, trilobite, unicorn, zebra, toad, triceratops



* i know some of these are technically not in the animal kingdom

5 faves
Uploaded on February 7, 2016