PDaaS (Personal Database as a Service)

Commands / HELP
Search result
Simple 3 steps

Inspiration

The meet with person is important for our life. However we can not remember all people. Sometimes there is a possibility that we miss a great opportunity because of it.

What it does

Our service can tell you who is in front of you easily even if your memory doesn't have anything. A few steps need to use it.

Input information you know before meeting to Slack. Our service gathers images from the internet and gives to Clarifai. Clarifai trains the machine learning model based on these images in advanced.

Send a picture and a tag data after you meet a person for the first time. Doing that, Clarifai trains the machine learning model.

Taking a picture using Spectacles glasses when you meet someone. You can get personal information about who is in front of you!

How I built it

Using Spectacles glasses to take a picture, and send it to storage. Clarifai trains the machine learning model with a tag. When you submit personal information, google custom search API gathers information to register train the model. When you meet someone as second time, the service can offer a person name as the inference of this model.

Using these components: Clarifai, Google Custom Search Engine, Spectacles, Lambda (AWS) etc

Challenges I ran into

Everything. It took time searching API specification because it's the first time for us to touch each services APIs: Clarifai API, Google Custom Search API, Spectacles API. Eventually, we were not able to good image information through Google Customer Search API. Improving Accuracy. In order to train, we need ten images at least. However it's not able to have accomplished for now. Sending image data from Spectacles glasses. We need to improve this part.

Accomplishments that I'm proud of

Anyway it works. Although it lacks some pieces.

What I learned

Clarifai training is so fast because of asynchronous. It's suitable for Machine Learning Continuous Integration.

What's next for PDaaS

How to get image information efficiently. Integrate with AR glasses to show the inference data. More sophisticated interface.