Fig. 3: Omniglot one-shot classification. | npj Artificial Intelligence

Fig. 3: Omniglot one-shot classification.

From: Learning from one and only one shot

Fig. 3

a shows the first two runs out of 20 total runs: the red outline marks one of 400 unit tasks, each consisting of 1 test and 20 training images. b displays the error-rate leaderboard: unlike all other comparison models, ours requires no background set for pretraining and achieves near-human performance.

Back to article page