Darknet

Darknet is an open source neural network framework written in C and CUDA. It supports CPU and GPU computation.

For more information see the Darknet project website.

For questions or issues please use the Google Group.

Installation

For general information on installation, you can take a look at darknet's website.

To install on MacOsx :

cd darknet
make

To install on Windows :

Follow this issue : pjreddie/darknet#721

Run

To execute the prediction's command, you will need to download:

• extraction.weights here (80MB)
• yolo.weights here (1GB)
• yolo.cfg here (3KB)
• yolov3.weights here (237MB)

Usage

First, create a ./data directory and store all your pictures into it. In that directory, create a subdirectory for every picture: for a picture named "voitures.png", you will store this file in a subdirectory named voitures. The path to the screenshot will be ./data/voitures/voitures.png.

Automatic labelization

To launch YOLO and store the coordinates of the items recognized:

• for windows

python python/rec_fct_slid_win.py mac data/voitures/ voitures

• for mac

python python/rec_fct_slid_win.py win data/voitures/ voitures

Visualize recognized areas

To check if the items automatically recognized are correct and visualize them, run drawing_rect.py as follows:

python python/drawing_rect.py --image data/voitures/voitures.png

Manual labelization

To labellize manually your pictures or screenshots, run click_and_get_coord.py as follows:

python python/click_and_get_coord.py data/voitures/voitures voitures