The source code includes a Variational Autoencoders project, which is an applied task: Web server for Searching similar images in a docker.
The VAE model was trained using the Fruits 360 dataset: A dataset of images containing fruits and vegetables. (https://www.kaggle.com/moltean/fruits)
The project has the following goals
- Design document and dataset description
- VAE Model implementation and training code providing by Jupiter Notebook
- Flask Server and HTML Frontend Client providing
- Training model from scratch by a dataset
- Compile docker and docker-compose for full architecture
- Fruits 360 dataset contains 90483 original sized and resised images
- Original Fruits 360 dataset description
- Original Fruits 360 resised dataset description
- Design drafts and the additional dataset description
- Model description and training log
- Dumped trained model
- Πhe main backend file
- Embedding of each image stored in './static/img/'
- This contains functions that will be used by the application to extract image embeddings.
- This keeps a record of all the images in './static/img/'
- File that contains information about the UI of the app. Teamplate images locate in '.static/templates/'
- Contains All the images that are used to construct the embeddings as well as images wich are using in the index.html file
- List of libraries used for this application
- Docker file
- Docker-compose file
You can run the app directly, or through docker and docker-compose.
To get the access to the web-front open http://127.0.0.1:5000/
gh repo clone galkinc/VAE-Searching-similar-images
cd VAE-Searching-similar-images
docker-compose build
docker-compose up
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent