People analysis in retail store FPS: 1.6
This is a kind of project that make analysis on people in store. For now, it's providing only people count data. Aim of this project is increase sales amount by providing useful data. You can see amount of people in specific field. This fields can be more likely hallway or something like that. You can also see points that people walk around in store. Started color is light blue , if a person stay same position more than one frame it gets dark.
Object tracking system can be implemented (I actually implemented DeepSort already but I did't use it because of Raspberry Pi limitations ) to provide diversity data. You can obtain the time that people walk around in specific field. For instance, people in field1 walk around in that field average 10 minutes.
- Flask
- Flask-Socketio
- OpenCv
- Numpy
- Pandas
- Shapely
- xlsxwriter
- Threading
- Python-Firebase
- Tensorflow lite run time interpreter
Firstly, the application runs in real-time. I use socketio to communicate to flask server in real-time. Basic idea updating data is sending socketio request to flask server every in 1 sec. So, this triggers the server to update data.
Server create a excel file by getting data from Firebase when clicked report button. Excel file contains counting data and some charts according to that data.
I trained on my own detection model in this project. Use Tensorlofw 2 object detection API . You can find training pipeline in train.ipynb file
To run this project
python3 app.py