A state-of-the-art detector of Boring hats. Developed at MIT to solve the global challenge (implicitly) posed by Elon Musk when he declared that 10 Boring hat owners will be awared a Boring tour. Given that 50,000 hats (current sale estimates) cover less than 0.001% of 7.3 billion heads on Earth, an automated method for Boring hat detection is needed.
The Boring detector is a fine-tuned deep neural network model based on a Keras implementation of a RetinaNet object detection architecture. The following are some results. The Boring Score is from 0 to 1. Anything above a score of 0.5 is considered Boring.
| Boring Detector | Hot Dog Detector | |
|---|---|---|
| Object localization | x | |
| Boring hat detection | x | |
| Hot dog classification | (in beta) | x |
If you would like an image or video of your own to be verified by the Boring detector, post it on Instagram or Twitter with hashtag #boringdetector (boring for lex). I'll then make links to processed versions of these images/videos here...
If you would like to run the network on your own machine, I'll post detailed instructions here soon.
Boring Detector uses Apache License 2.0. See LICENSE for more details.