This is a personal implementation of YOLO v5 by Pi.

Dataset : https://drive.google.com/file/d/11q37nAXEVgagxyj2LjgkXfc2LzNqNbmo/view?usp=sharing

Python >= 3.6.0 required with all requirements.txt dependencies installed:

To test the repository, run these commands first.

$ git clone https://github.com/Pi-31415/yolo-personal
$ sudo apt install python3-pip
$ sudo -H pip3 install --upgrade pip
$ sudo -H pip2 install --upgrade pip
$ cd yolo-personal
$ pip3 install -r requirements.txt

Then, place the image you want to test into the yolov5 folder. In this case, it is called 3.png.

Then run the detect command.

$ python3 detect.py --weights runs/train/exp10/weights/last.pt --img 640 --conf 0.25 --source 'a.png'

Note : 5 = first train 7 = with new ojects 10 = latest train

The result will be in runs/detect/exp (read the terminal output for exact location)

To train the model, train with

python3 train.py --img 640 --batch 16 --epochs 2500 --data sciroc.yaml --weights yolov5s.pt --cache --nosave

To get images from ros node, run

rosrun image_view image_saver image:="/xtion/rgb/image_rect_color"

To copy exchange folder contents to the docker, run

cp ../exchange/* ./

To run simulation, run

roslaunch tiago_998_gazebo tiago_navigation.launch

Notes

• There are some warnings (WARNING: Value for scheme.headers does not match.) when installing required python packages, but does not affect the simulation at all.
• torch>=1.7.0 (831.4 MB) will take a while to download everytime docker container is started.