Want to take your software engineering career to the next level? Join the mailing list for career tips & advice Click here


GoogLeNet on iOS demo

Subscribe to updates I use MetalDetector

Statistics on MetalDetector

Number of watchers on Github 43
Number of open issues 0
Average time to close an issue 2 days
Main language Metal
Average time to merge a PR less than a minute
Open pull requests 0+
Closed pull requests 0+
Last commit almost 2 years ago
Repo Created almost 5 years ago
Repo Last Updated 4 months ago
Size 49.7 MB
Organization / Authorkrasin
Page Updated
Do you use MetalDetector? Leave a review!
View MetalDetector activity
View on github
Book a Mock Interview With Me (Silicon Valley Engineering Leader, 100s of interviews conducted)
Software engineers: It's time to get promoted. Starting NOW! Subscribe to my mailing list and I will equip you with tools, tips and actionable advice to grow in your career.
Evaluating MetalDetector for your project? Score Explanation
Commits Score (?)
Issues & PR Score (?)


GoogLeNet on iOS demo.

Note: this demo was implemented before TensorFlow, PyTorch or Core ML existed. It's preserved for historical purposes only.

This is a basic demo of GoogLeNet from Caffe Zoo working on iOS. It takes the input from the camera and prints the top prediction. The speed is 1 FPS on iPhone 6S, and 3 seconds per frame on iPhone 6.

Two known bugs:

  • You need to hold the phone so that the round button is on the right. Otherwise, the network will get a rotated image, and the classification will likely miss.

  • There's a rounding bug right now (likely, in Convolution layers). While the network gives answers which are within ~3% of Caffe output, it's slightly worse than a real GoogLeNet taken from the Caffe Zoo.

Also, please, be aware that ImageNet classes are weird, not so many real world things could be detected (but it knows about 300 breeds of dogs, whee!)

The code was generated with Goffe, Caffe-to-{Swift+Metal} generator.

MetalDetector list of languages used
Other projects in Metal
Powered by Autocode - Instant Webhooks, Scripts and APIs
Autocode logo wordmark