Get Started

In Anaconda Prompt or the Terminal

cd your_directory

Creating new virtual environment

Create new folder name: "sign_project" on your directory. Put your own environment name instead of "envName" or you can you this name.

mkdir sign_project
cd sign_project
python -m venv envName

Activate your environment

envName\Scripts\activate.bat

Using Git clone to set up the model tools

how to set up git click >>

git clone https://github.com/DreamPTK/yolov5.git

Open Visual Studio Code (VScode)

code .

In Visual Studio Code (VScode)

Open detect.ipynb on yolov5 folder and

Open New Terminal in Vscode (command Prompt)

cd yolov5
pip install -qr requirements.txt

Import module and set up tools

In detect.ipynb

import os
import utils
display = utils.notebook_init()

Detection Sign image

The detect.py run ThTaxSign model on all images source of data/images folder and save the result to output/exp folders.

This case study have two images source on data/images folder.

# use ThTaxSign model to detect 
!python detect.py --img 640 --weights ThTaxSigns/data/models.pt --conf 0.3  --source ThTaxSigns/data/images --project ThTaxSigns/data/output --save-crop

--img: inference size (height, width)

--weights: model path or triton URL

--conf: confidence threshold

--source: file/dir/URL/glob/screen/webcam (source of the image)

--project: save results to project/name

--save-crop: save cropped prediction boxes

Display output

# Display image

import glob
from IPython.display import Image, display

for imageName in glob.glob('ThTaxSigns/data/output/exp/*.jpg'): 
    display(Image(filename=imageName))
    print("\n")

123

Display crop output

This tool will separatly crop the sign image output to several image files.

# Display all crop images

for imageName in glob.glob('ThTaxSigns/data/output/exp/crops/unidentified_signs/*.jpg'): 
    display(Image(filename=imageName,width=300))
    print("\n")

123

Finally, you can see these crop image outputs on output/exp/crops/unidentified_signs folder.