/glasses

Model built using pretraining restnet50 model in pytorch, from the data collected using google-images-download. Also contains visualization techniques for the prediction using saliency maps.

Primary LanguageJupyter Notebook

Glasses

Checks if the user is wearing any glasses or not

Working Image with glasses Working Image without glasses

Technologies used

  • PyTorch
  • Flask

Main Things Learnt

  • Creation of own Datasets
  • Use of Pretrained Models
  • Deploying Pytorch models over Flask server

Features to add

  • Add webcam support to take pictures on the page
  • Asynchronously add real time detection and witty comments Thanks to @memahesh
  • Make homepage more presentable

Installation / Testing

  • Clone the repository git clone https://github.com/Mianto/glasses.git and change the folder to Glasses.
  • Install the dependencies pip install requirements.txt
  • Change folder to glasses and run the flask app using python app.py

Files

  • glasses.ipynb used for training the model
  • saliency.ipynb gives the saliency map for the image