Yolov5 object detection model deployment using flask

This repo contains example apps for exposing the yolo5 with custom object detection model(sea.pt)

Web app

Simple app consisting of a form where you can upload an image, and see the inference result of the model in the browser. Run:

$ python webapp.py --port 5000

or

$ python3 webapp.py --port 5000

then visit http://localhost:5000/ in your browser:

The example json result

[{'xmin': 47.1176986694, 
'ymin': 221.4556274414, 
'xmax': 401.8820495605, 
'ymax': 557.4293823242, 
      'confidence': 0.7438420057, 
      'class': 1, 
      'name': 'stingray'}, 

Run & Develop locally

Run locally and dev with anaconda :

• conda create -n python37 python=3.7
• conda activate python37
• git clone https://github.com/serversuck/yolov5-flask.git
• (python37) $ pip install -r requirements.txt
• (python37) $ python webapp.py --port 5000

YoloV5 model training by

mr sorapong somsorn sorapon@varee.ac.th

reference

• https://github.com/ultralytics/yolov5
• https://github.com/jzhang533/yolov5-flask (this repo was forked from here)
• https://github.com/avinassh/pytorch-flask-api-heroku