This project aims to use deep learning (transfer learning) to identify the style of furniture given an image.
More details:
-
A Resnet34 CNN image classification model is trained with images of Mid-century modern, Rustic, Arts and Crafts, Traditional styles of furniture. (see Notebooks for training)
-
pytorch_utils is a separate python module that is used during training and can be used with other pytorch projects. It consists of utils for callbacks, data processing and model building.
-
Webapp is a Flask based webapp. User can upload an image file or an URL. The webapp predicts the probabilities of each class based on a checkpoint model. It allows user to update the class label if the prediction is wrong. Every image upload, predictions and user input for correct label are recorded into a database.
-
Data for training was obtained from Google Images. DEMO hosted on google cloud platform
| Home Page | Prediction/Feedback Page |
|---|---|
 |
 |