Self Defense Trainer using opencv & GCloud AutoML Vision

• 

  Front of app contains a brief explanation of how the app works. User may also choose their difficulty here; harder difficulty = less time.

• 

  This is feedback on whether the user gave the proper response.

• 

  The program gives the user a pose prompt to respond to. (Meant to be used with full body of user within the frame.

Inspiration

We intend to empower people especially women not only mentally, but physically as well through teaching self defense so that they may feel confident and safe to protect themselves and their loved ones if the time comes.

What it does

This program teaches the user self defense by training them to respond to indicators of physical threats that are most likely to be encountered within their daily lives.

For example: A right punch from a possible assailant can be countered by blocking left or weaving under.

How we built it

Using the Google Cloud AutoML API to train a model of self-defense move recognition as well as a basic Python GUI for portability on any PC with a webcam, we built a program that prompts users to train their self defense instincts.

Challenges we ran into

• Communicating with the cloud to access our custom trained model.
• Utilizing the output of our prediction method to be able to organically interact with the purpose of our program

Accomplishments that we're proud of

What we learned

• Training the model is half the battle

What's next for Self Defense Trainer using opencv & GCloud AutoML Vision

• Training a new model to determine if a series of attacks are able to stun or subdue a potential assailant
• Keeping track of user's progress to be able to motivate them to keep training.

Built With

• automl
• google-cloud
• machine-learning
• opencv
• python