This is a tool to generate synthetic images where provided images of objects are scattered over a white background or over optionally provided background images.
As this is not yet published in PyPI, manual installation is required.
-
Clone this repository to local
git clone git@github.com:schapagain/scatter-yolo.git
-
Create a new virtual environment and install
flitconda create -n scatter-yolo flit
-
In the root project directory run:
flit install -
Now you can start using the tool with. To check out all the generation options, run:
scatter-yolo generate -h
scatter-yolo generate samples/images/fish samples/images/turtles
scatter-yolo generate samples/images/fish samples/images/turtles -b samples/images/ocean_bed
scatter-yolo generate samples/images/fish samples/images/turtles -b samples/images/ocean_bed -c 2
|
scatter-yolo generate samples/images/fish samples/images/turtles -b samples/images/ocean_bed -rat 0.2,0.8
scatter-yolo generate samples/images/fish samples/images/turtles -b samples/images/ocean_bed -min 10 -max 20
* The exact count of scattered objecs cannot be set, and only min and max counts are accepted. The max parameter is a true maximum, however there might be fewer objects than the min parameters (if there is no space to place enough objects in the image, for example). This is to vary the number of objects in each image when doing bulk generation.
scatter-yolo generate samples/images/fish samples/images/turtles -b samples/images/ocean_bed -min 10 -max 20 -os 50
scatter-yolo generate samples/images/fish samples/images/turtles -b samples/images/ocean_bed -sh circ
scatter-yolo generate samples/images/fish samples/images/turtles -b samples/images/ocean_bed -cx 0.8
scatter-yolo generate samples/images/fish samples/images/turtles -b samples/images/ocean_bed -anim yes
