TinyTags

TinyTags allows you to easily search for images within the 80 million tiny images dataset provided by MIT and NYU. Our TinyTags model learns to categorize the images using a predefined set of tags, producing an accuracy estimate for each category prediction. The tags are obtained from the smaller 60,000 CIFAR-10 dataset that provides 32x32 color images labeled as "deer", "automobile", "truck", "cat", "dog", "bird", "airplane", "ship", "frog", and "horse."