My colleagues and I have decided to publish Babbly Android app to help people make most out of the isolation by learning a new language.
Babbly is one of our side projects we have been developing for the past few months. Our team has a lot of expertise in Computer Vision and we have always wanted to test if in-the-cloud real-time inference is possible with AI. Babbly can process up to 5 pictures per second which demonstrates this approach is more than viable. The main advantages of the in-cloud model are seamless AI model updates, reduced battery consumption, and much lower restrictions on the model size which results in better accuracy.
The model was trained on ImageNet and OpenImages public datasets which are not perfect for this kind of computer vision task, so the accuracy is not stellar. We plan to work on this issue in the future :)
We realize this app doesn't offer much feature-wise, but we think any form of fun/knowledge can be helpful during this pandemic. And yeah, any kind of feedback is more than welcome :)
Babbly is one of our side projects we have been developing for the past few months. Our team has a lot of expertise in Computer Vision and we have always wanted to test if in-the-cloud real-time inference is possible with AI. Babbly can process up to 5 pictures per second which demonstrates this approach is more than viable. The main advantages of the in-cloud model are seamless AI model updates, reduced battery consumption, and much lower restrictions on the model size which results in better accuracy.
The model was trained on ImageNet and OpenImages public datasets which are not perfect for this kind of computer vision task, so the accuracy is not stellar. We plan to work on this issue in the future :)
We realize this app doesn't offer much feature-wise, but we think any form of fun/knowledge can be helpful during this pandemic. And yeah, any kind of feedback is more than welcome :)