This project utilises unsplash API to fetch the images according the query (word) searched. After getting the results we then went ahead and cluster them according to the similarity and shown the images on the web page with the help of streamlit. The KMean model was also used to help us cluster the images. The app has been deployed on Streamlite since it offers easy and smooth presentation of results.
Note: Deployed version of the web pages Here
- streamlit for creating the web app interface.
- requests to make HTTP requests to fetch images from an API.
- TfidfVectorizer from sklearn.feature_extraction.text for converting text descriptions into TF-IDF features for text-based clustering
- KMeans from sklearn.cluster for clustering images based on their pixel data or associated text descriptions.
- Other libraries like
osfor operating system interactions,PIL.Image, numpyfor image processing, andpytesseractfor text extraction.
Please note that these packages should be installed first, as per requirements.txt
To run the project locally, there is a need to have Visual Studio Code (vs code) installed on your PC:
- VS Code: It is a source-code editor made by Microsoft with the Electron Framework, for Windows, Linux, and macOS.
- Clone the project
git clone https://github.com/UmuhireJessie/image-clustering.git
- Open the project with vs code
cd image-clustering
code .- Install the required dependencies
pip install -r requirements.txt- Run the project
streamlit run image.py- Use the link printed in the terminal to visualise the app. (Usually
http://127.0.0.1:8501/)
- The app has used existing APIs (unspash API) for searching the images so one may be required to sign up for Unsplash developer account and create project in order to generate an ACCESS KEY. That is if they want to test the app locally.
Jessie Umuhire Umutesi