Harvest_estimator

A Django based web project that uses custom YOLO model, trained on google open images, to detect and count number of fruits(Apple, Mango, Orange, Pomegranate, tomato) in an image.

Requirements:

python >3.4 with pip

make sure the PATH is correctly modified for pip to work,

	e.g. C:\Python34\;C:\Python34\Scripts;

setuptools

pip install setuptools

opencv 3.4

pip install opencv-python

django

pip install django

Setting up the project

• grab a copy of the project - unzip it

• go to project root fruitDetect

• run python manage.py runserver

The Project should be live and accessible at http://localhost:8000/scan

Project Structure

submodule parseImage handles the scanning

submodule fruitDetect is teh main application and has teh settings.py for Django project.

static directory contains the images which are generated after scanning

yolo-fruits is the custom trained model to detect below fruits:

Mango
Apple
Orange
Pomegranate
Tomato

yolo-coco(not being used) is the pretrained yolo model that detects ovber 80 classes folow teh below url to know more.

to perform a generic object detection change the yolo-scan function in views.py in parseImage submodule. 

yolo-structure

.names file - contains serialised names of classes trained on
.cfg file - contains the yolo configuration that determine the whole model, its layers and other details. !! Please change only if you know what you are doing !!
.weights file - contain the weights that are achieved by training Yolo of a set of classes mentioned in obj.names (this file is huge and has to be trained by yourself )