machinevisionbeans / streetdensityai Goto Github PK

This YoloV5 based model is fit to detect people and different types of land vehicles,
and displaying their density on a fitted map, according to their coordinates and detected labels.

Features:

• Multiple Objects Detection
• Trained on 3,000 street view Images
• Exports Fitted and adjustable Maps
• Calculates a density score according to image detected labels

• in order to plot the map: pip install mplleaflet
• to use YoloV5, see: Yolov5_requirements

Object Detection:

identifies people and land vehicles in your images:

python src/yolov5/detect.py --source  <path to images folder> --project <output path>
--name <output folder name> --save-txt --conf 0.3

running this action will save your images with the anchor boxes around objects that the model found:

(if you don't want to save the labeled images, just add --nosave to the command above) in addition, it will save the detected object labels for each image.

Plotting a fitted map:

display the density on a fitted map (requires a .csv file)

python src/steetdensityai.py --labels <labels path that were created after the images detection>
--coordinates <path-to-csv/file.csv>  --images <path to images folder>
--img-per-cord 1 --output <output path>

• csv requires 2 columns to display the coordinates named: "longitude" and "latitude"
• the code asumes that the coordinates are sorted by the image's name.
• If you have multiple images per coordinate (for example if you have a 360 view, divided to 4 images), you can set the number of images per coordinate with : --img-per-cord <integer of images per coordinate >

# detect objects: 
python src/yolov5/detect.py --source example/images --project example/images --name detected_images --save-txt --conf 0.4


# creates a label folder in example/images/detected_images named "labels"
# saves the images with the newly found objects anchor, and each image labels 


#plot desnity map
python src/steetdensityai.py --labels example/images/detected_images/labels --coordinates example/coordinates.csv  --images example/images --img-per-cord 4 --output example/images

 # will save the map.html file to example/images