👉 Our model is now live here!
- ALLIENNE Maxime
- DERONNE Jules
- DRUON Constant
- HORNAIN Imane
- PROUFF Benjamin
For this project, we decided to build a model to detect whether a person is attractive or not. But while building our model, we realized a problem. The attractiveness factor is subjective. This subjectivity implies a model with poor performance.
So, to illustrate the problem, we built a second, identical model that detects if someone is smiling. We've choose smile, because it's a objective attribute. And as expected, this model gets much better results.
You can find these 2 models in :
- model/attractiveness.ipynb
- model/smiling.ipynb
In order to run model files, please create a folder input inside /model. Then download the CelebA dataset, unzip it and add it inside this new folder. Then you're ready to go !
Finally, we've decided to integrate our model into a website. It's more fun and convenient to classify images in that way!
To do this, we converted the .h5 template into a Tensorflow.js compatible template. Then we created a React.js project that allows to upload a picture (locally only). When the user adds a photo, we predict if the person is smiling or not.
You can test our model on this this link.
