Matthieu Hanaania
This project is a django website associated with a object detection deep learning model.
Install tensorflow pip install tensorflow==2.14.0
Download the tensorflow github from git clone --depth 1 https://github.com/tensorflow/models
copy the setup.py file from models/research/object_detection/packages/tf2/setup.py to models/research/
In the list of required packages, add 'PyYAML==5.3.1',
and then execute python -m pip install .
Then install django pip install Django
Dowloading_a_model_whithout_retrain.ipynb
For this project, I use an existing and already trained model. (ssd_mobilenet_v2_320x320_coco17_tpu-8 )
The model is saved as a folder with the weight details and the :configuration file.
It has been train of the coco dataset : a large dataset with 90+ categories.
By loading the model, it is possible to send it an image, and it will detect different objects, based on the categories it has been trained on. It will frame the detected objects with boxes and attribute each boxe with an index. By knowing that the bottles are the 44th index. We can keep only the boxes associated with this index.
With the object_detection.utils library it is possible to put boxes on the image, and save it with the cv2.imrite function
The Django architecture is firstly composed of a projet "Website" with its configuration files in the "mysite" folder. Most of the content is automatically. in the "urls" file, we define all the urls link of the application.
It is in the "plasticDetection" application that we develop the plastic-detection website.
The website is divided in 2 pages. The image submission for AI plastic detection, and the global map page
Everything is based on views in the "views.py" file. It define all the functions that can interact with the HTML website. The "urls.py" file redirect every path to a view.
Moreover, Django contains a sql database, defined in the "models.py file"
To run the website, execute the command py Website\manage.py runserver
and then go to the link http://127.0.0.1:8000/plastic-detection/
The Django architecture is divited into website
This page calls the model, that detect plastic bottle in the user-submitted picture.
You can submit an image, and call the AI model. That will frame some plastic waste in the picture.
If an image contains geographical information ( longitute and latitude ) they will be print on the web page
When an image is submitted, its position is stored in a sql database that is displayed in this page.
