Tired of using image recognition APIs to recognize images? Get ready to break out of the mold, with the awkwardly titled Speakifai.
What it does
Speakifai uses the Clarifai image recognition API to recognize voices.
How I built it
I built a database of audio recordings of myself, converted them to spectrographs, normalized them, and trained Clarifai to associate them with words.
Challenges I ran into
Jumped around between several python UI toolkits before finally settling on a web-based approach. Data normalization was difficult. Had trouble with the fitting.
Accomplishments that I'm proud of
It performs admirably on some of my trials, correctly identifying speech under varying conditions with a reasonable accuracy rate. With some work, it could be greatly improved.
What I learned
More data would have been very helpful. I would have loved to have had a more graphical presentation.
What's next for Speakifai
More data and a better UI.