Click here to download the executable file (Windows users)
The easiest option to run the code is by downloading the folder from Github and double-click the .exe file.
You can find a folder with the executable (application) and also with the source code (.py).
If you would like to view the Python code and perhaps do some changes, you can download the source code (.py) file. I did the programming-testing-debugging using the highly recommendable Spyder IDE.
You have to open and execute the source code (.py) in a free Python-based IDE, for example Spyder or Pycharm.
Before you can use it, you need to install some Python modules using the command terminal:
You have to open the source code (.py) in a free Python-based IDE, for example Thonny. I had no problems using the software on a Raspberry Pi 4B with 4GB RAM (€50) using Linux, after installing the following Python packages using the terminal.
First of all: install pip for linux
apt install python3-pip
To install the pandas:
pip install pandas
To install the opencv module
pip install opencv-python
Using my ultra-cheap Raspberry Pi I noticed that the program didn't want to initiate because opencv2 was not entirely well installed. I found some help on my favorite website (stackoverflow) and found that it's possible to solve this error while executing the following code in the Linux command shell:
sudo apt-get install libcblas-dev
sudo apt-get install libhdf5-dev
sudo apt-get install libhdf5-serial-dev
sudo apt-get install libatlas-base-dev
sudo apt-get install libjasper-dev
sudo apt-get install libqtgui4
sudo apt-get install libqt4-test
During debugging I noticed that Windows uses typically " \ " and Linux uses "/". This means that this needs to be changed because the software will give a bug.
Manual here Old version tutorial New short tutorial New long tutorial
-
Convert the training images: I used for example 804x804px images. In order to resize the images I used ImageJ or Fiji. I used process => Batch => Convert in order to convert the original 1608x1608px tiff files to a 804x804px jpeg format.
-
Install LabelImg and draw boxes around the tracks in the images
-
Compress the labelled files and their .txt files in one .zip file and place this file images.zip in the yolov3 folder on your google Drive.
-
Execute the Google Colab notebook for some hours. Aim for 2000-3000 iterations. The iteration speed depends on the size of your images. From my experience it seems that from then on, the DNN is overtrained.