cvlayer
OpenCV Layer Helper
Overview
When performing image-based machine learning/inference, many situations arise that require pre-/post-processing appropriate for the domain environment.
A commonly used library today is OpenCV.
Image processing with OpenCV requires insight into the image and the domain.
However, gaining these insights often requires doing something far from the essence. (Visualization, parameter modification, library conflicts, assertions, etc...)
Here are some problems that occurred to me:
...
Features
...
Installation
Install cvlayer:
pip install cvlayer
Install cvlayer with opencv-python:
pip install cvlayer[opencv]
Install cvlayer with opencv-python-headless:
pip install cvlayer[headless]
Usage
CvLayer
Just inherit cvlayer.CvLayer.
from cvlayer import CvLayer
class YourApp(CvLayer):
def func(self, image):
self.cvl_cvt_color_bgr2hsv(image)
CvWindow
Just inherit cvlayer.CvWindow.
from cvlayer import CvWindow
from numpy.typing import NDArray
class YourApp(CvWindow):
def on_frame(self, image: NDArray) -> NDArray:
# TODO: Implement image processing logic ...
return image
YourApp("/path/to/video/file.mp4").run()
Example
The following sample is a Perspective Transform example:
from sys import argv, stderr
from sys import exit as sys_exit
from typing import List, Optional
from cvlayer import CvLayer, CvMixin, CvWindow
from cvlayer.typing import PointI
from cvlayer.palette.basic import RED
from numpy.typing import NDArray
class YourApp(CvWindow, CvMixin, CvLayer):
_points: List[PointI]
def __init__(self, source: str, destination: Optional[str] = None):
super().__init__(source, destination)
left_top = 930, 2750
left_bottom = 846, 3098
right_top = 1091, 2750
right_bottom = 1361, 3098
self._points = [left_top, left_bottom, right_top, right_bottom]
self._scale = 2, 4
def on_frame(self, image: NDArray) -> NDArray:
with self.layer("select-roi") as layer:
self.roi = layer.param("roi").build_select_roi().value
layer.frame = layer.prev_frame
with self.layer("perspective-points") as layer:
points = layer.param("pp").build_select_points(self._points).value
canvas = layer.prev_frame.copy()
for p in points:
self.cvl_draw_point(canvas, p, color=RED)
layer.frame = canvas
with self.layer("perspective-transform") as layer:
sw = layer.param("scale-width").build_uint(self._scale[0]).value
sh = layer.param("scale-height").build_uint(self._scale[1]).value
xs = list(map(lambda point: point[0], self._points))
ys = list(map(lambda point: point[1], self._points))
x1, x2 = min(xs), max(xs)
y1, y2 = min(ys), max(ys)
width, height = abs(x2 - x1) * sw, abs(y2 - y1) * sh
roi = 0, 0, width, height
m = self.cvl_get_perspective_transform_with_quadrilateral(
left_top=points[0],
left_bottom=points[1],
right_top=points[2],
right_bottom=points[3],
destination_roi=roi,
)
layer.frame = self.cvl_warp_perspective(image, m, (width, height))
with self.layer("hsv") as layer:
layer.frame = hsv = self.cvl_cvt_color_bgr2hsv(layer.prev_frame)
with self.layer("hsv-h") as layer:
layer.frame = h = hsv[:, :, 0]
with self.layer("hsv-s") as layer:
layer.frame = s = hsv[:, :, 1]
with self.layer("hsv-v") as layer:
layer.frame = v = hsv[:, :, 2]
assert h is not None
assert s is not None
assert v is not None
self.cvm_gaussian_blur("v-blur", (3, 19), 0.0, 7.0)
self.cvm_threshold_binary("v-thresh", 230)
return self.last_frame
def test_main(*args) -> None:
source = args[1]
destination = args[2] if len(args) >= 3 else None
try:
app = YourApp(source, destination)
app.run()
except Exception as e:
print(e, file=stderr)
sys_exit(1)
else:
sys_exit(0)
if __name__ == "__main__":
test_main(*argv)
License
See the LICENSE file for details. In summary, cvlayer is licensed under the MIT license.
1 maintainer
Metadata
1 maintainer
Release files for cvlayer 0.32.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| cvlayer-0.32.0-py3-none-any.whl | Python 3 | none | any | Details |
Release files / cvlayer-0.32.0-py3-none-any.whl
| Download URL | cvlayer-0.32.0-py3-none-any.whl |
|---|---|
| Size | 197.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
0010a6dede5885c8c5a99b7ba5c2db713a8871dc62b2c20fed4a300d5b171052
|
|
BLAKE2b-256 checksum How to use checksums |
12fe0f7f47f9581f3576f49e285ae676595f9f7e4d2c8072f02ebf45f3fd28f5
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/5.1.1 CPython/3.9.19
|
Security for cvlayer 0.32.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
