PriceWaterHouseCooper's survey says that about 5 - 15% of all IT projects ($50 - $150 million loss) face setbacks in some form or the other. With the advancements in project management and scheduling tools, and the fact that 90% of a Project Manager's job is communication, we can surmise that it boils down to handling teams and dynamics. As humans, we are not perfect - we are often biased by various internal or external factors that we have no control over which can affect the way we make decisions and how we act. If we could instead, use a machine to assist us with some aspects of these decisions - and they are useful for at least improving delays by 0.1%, we can stand to save at least $1 million, which, is a large sum.

What it does

This network can identify a person's emotion (one of seven currently trained) through a photo or video in real time

How we built it

Using Azure ML and a data set of faces that portray one of 7 emotions, we built a neural net to identify the emotions in real time through a web service. We also have a python script to test the endpoint (manual face detection) and a C# script to capture faces and transmit them (automatic face detection)

Challenges we ran into

Finding a good dataset, parameter tuning, business opportunities

Accomplishments that we're proud of

We got 78% accuracy on our set - almost state of the art. We could also use this net in real time emotion recognition.

What we learned

How powerful neural networks are for multi-class image classification based on large feature vectors, and that emotions are not so easy to recognize.

What's next for aeiNN

By applying this net to real time project and team monitoring systems, we can semi (and in the future) fully automate project management by accounting for tangible and intangible components. This is just a demo of how powerful this tool can be for team handling and team load distribution.

Built With

Share this project: