- created at Enghacks 2019 in 48 hours by Penny Ji, Andrew Mourcos, Aidan Philpott and Robert Toyonaga.
Pierre helps people to overcome public speaking anxiety by improving their presentation skills and giving them confidence.
Pierre is a presentation tool that makes use of natural language processing as well as computer vision heuristics in order to make you a better speaker. In real time, Pierre is able to catch utterances (ex: 'like', 'um', 'literally', ...) and visual nervous ticks (swaying, tapping, ...) which distract from your presentation. As Pierre catches these faults, he alerts the speaker in real-time (while practicing) via smartphone notifications and can display analytics afterwards.
PocketSphinx was the framework used for live NLP, OpenCV was used to develop image processing heuristics for "tick detection" and Stdlib was used for notifying the speaker during the presentation and as a database for analytics.
Proper user research was conducted in order to evaluate the effectiveness of Pierre as a presentation tool. Additionally, several people at the event were interviewed in order to validate the utility of our solution.
We will be conducting more user research to help us with our product direction. The next big update will be the addition of a GUI that allows users to view their live data as they are presenting and also compare their recent sessions. Additionally, we will be doing some research to develop more robust tick/utterance detection algorithms.
Clone/download this repository and make sure you have Python3 (atleast 3.4) installed.
Use pip to collect the necessary dependancies:
pip3 install -r requirements.txt
Run main.py and give it a shot!