Team-RoomNumber404/backend_server

Django Object Detection With YoloV5

Demo of the WebApp

DjangoObjdetection.mov

This is a project to demonstrate easy integration of YoloV5 in Django WebApp and not a full-fledged production ready app though can be scaled to work as one.

Features of the WebApp:

• Create/Edit ImageSets.
• Upload multiple images with dropzonejs to the selected ImageSet.
• Convert uploaded image size to 640 x 640. (For faster detection)
• Upload/update a custom pre-trained model.(If you have offline files of a model)
• YoloV5 models will download upon selection. (Active internet connection required for this step.)
• Detect object on an image with YoloV5/custom pre-trained model.

Note

a default.png in media is required for user-profile. Create media folder and add a 'default.png' file here.

Steps to use locally.

clone the repo locally

create virtual env

# install dependencies
pip install -r requirements.txt

# migrate
python manage.py migrate

# create super user
python manage.py createsuperuser # (it may show an error page if no 'default.png' in media folder. See note above.)

# run
python manage.py runserver

login
# Login at the web address 127.0.0.1:8000 using the superuser credentials.

Create ImageSet
# create an ImageSet first and then upload images into the ImageSet from ImageSet detail page.

# On images list page click on detect object.

# select a YoloV5 model
# the YoloV5 dependencies and pre-trained model will start downloading.

Apps:

• Detectobj
• images
• modelmanager
• users

Main dependencies

• Django
• YoloV5
• Pytorch
• Torchvision
• Tensorboard
• Pillow
• django-crispy-forms
• django-cleanup

Javascript library

• dropzonejs
• ekko-lightbox

Django starter template used

DjangoAdvancedBoilerplate