Virtual Walks in Google Street View

August 9, 2020 Sunil Ghimire

Due to the pandemic situation of COVID-19, we are now facing our freedom to walk openly on the street is being restricted in pursuit of the community’s well-being. So for the good health of the citizen, the government has planned to lock the border and in most countries flying is completely prohibited. In such a case, people are requested to go out for certain situations like buying the grocery. Today in this tutorial, we are going to demonstrate a virtual walks project.

So, this project is based on Pose Estimation Models along with Long Short Term Memory (LSTM). The PoseNet model was used for pose estimation, while an LSTM model has been developed using TensorFlow 2.0 for the behavior detection portion. Thus, with the aid of Google Street View, the virtual walks project is able to mimic walking around the street all around the world.

Technologies used in virtual walks program:

1. Python 3.7.3

2. TensorFlow 2.0

3. Selenium

Step involved to run this program:

1. Download the repository and unzip the file

2. Install all the dependencies

3. Download and install Firefox and Geckkodriver

4. Run execute.py

5. Finally, enjoy virtual walks program

Note: Webcam and Python 3.7 is needed to use this program to avoid issues and version conflicts. Also, you need to install tensorflowjs to run ResNet.