Landmark Recognition

An end-to-end application to predict landmarks from images, retrieve geolocation details, and visualize results on an interactive map. Check Blog here.

Features

• Simple repsonsive UI.
• It will give you the full address of Landmark
• It will provide you the Latitude & Longitude of predicted landmark.
• It will plot the predicted landmark on the Map.

Model & Data Source

• Leveraged a Pretrained TensorFlow-Hub Model (landmarks_classifier_asia_V1/1) for Landmark Classification
• Supports roughly 98,961 landmark classes covering Asia’s most iconic sites.
• This model was trained on Google Landmarks Dataset V2.

Usage

• Clone this repository.

• Open CMD in working directory.

• Run following command.

  pip install -r requirements.txt
  

• LM_Detection.py is the main Python file of Streamlit Web-Application.

• To run app, write following command in CMD. or use any IDE.

  streamlit run lm_recog.py
  

• For more explanation of this project see the tutorial on Machine Learning Hub YouTube channel.

Screenshots