A simple web application to classify images into categories described.
This is an interactive convolutional neural network on a website. The aim of this project is to feed in an image and run it through a CNN at the backend to display the best possible classification for the uploaded image.
The web app is segregated into 2 parts: Front-end and Back-end. Front-end runs on a microservice with routes on Flask and the backend runs on Tensorflow and Keras.
The application runs on Flask as the primary web interface and the CNN's are implemented on a Tensorflow backend. Bokeh is used to give a visual description of the classification.
Setup:
I am using a virtual environment to manage my dependancies.
$ mkdir ~/venvs/
$ virtualenv ~/venvs/appname
... # some output messages
$ source ~/venvs/das/bin/appname
(das) $ which pip /home/das/venvs/bin/pip
Then, pip3 install these dependancies,
- hdf5
- imageio
- flask
- Bokeh
- TensorFlow
- Keras
Once, all the dependancies are installed and cloned, run FLASK_APP=server.py flask run
The sample dataset I have used to get the images and train them are on sample aws s3 buckets. I imported them locally and trained the models on it.
Project status until I resolve deployment:
After uploading your image, simply click on Submit to predict the image.
