potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-29 23:33:05 +04:00
Code Packages Releases Wiki Activity

Merge pull request #28588 from abhishek-gola:ORT_GPU_wrapper

Browse Source 
Added OnnxRuntime GPU wrapper #28588

### Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [x] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
      Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
This commit is contained in:
Abhishek Gola
2026-04-06 18:10:58 +05:30
committed by GitHub
parent a49a293d3c
commit 2ec6a6bb65
19 changed files with 402 additions and 135 deletions
Download Patch File Download Diff File Expand all files Collapse all files
samples/dnn/alpha_matting.py
+5 -5
View File
@@ -129,11 +129,13 @@ def apply_modnet(args, model, image):
image, args.scale, (args.width, args.height), args.mean, swapRB=args.rgb
)
model.setInput(inp)
t0 = cv.getTickCount()
out = model.forward()
t = (cv.getTickCount() - t0) / cv.getTickFrequency()
alpha_mask = postprocess_output(image, out)
alpha_3ch = cv.merge([alpha_mask / 255.0, alpha_mask / 255.0, alpha_mask / 255.0])
composite = (image.astype(np.float32) * alpha_3ch).astype(np.uint8)
return alpha_mask, composite
return alpha_mask, composite, t
def main(func_args=None):
@@ -156,10 +158,8 @@ def main(func_args=None):
args.model = findModel(args.model, args.sha1)
net = loadModel(args, engine)
alpha_mask, composite = apply_modnet(args, net, image)
t, _ = net.getPerfProfile()
label = "Inference time: %.2f ms" % (t * 1000.0 / cv.getTickFrequency())
alpha_mask, composite, t = apply_modnet(args, net, image)
label = "Inference time: %.2f ms" % (t * 1000.0)
draw_label(image, label, (0, 255, 0))
draw_label(alpha_mask, label, (255, 255, 255))
samples/dnn/classification.py
+3 -2
View File
@@ -135,7 +135,9 @@ def main(func_args=None):
# Run a model
net.setInput(blob)
t0 = cv.getTickCount()
out = net.forward()
t = (cv.getTickCount() - t0) / cv.getTickFrequency()
(h, w, _) = frame.shape
roi_rows = min(300, h)
@@ -143,8 +145,7 @@ def main(func_args=None):
frame[:roi_rows,:roi_cols,:] >>= 1
# Put efficiency information.
t, _ = net.getPerfProfile()
label = 'Inference time: %.1f ms' % (t * 1000.0 / cv.getTickFrequency())
label = 'Inference time: %.1f ms' % (t * 1000.0)
cv.putText(frame, label, (15, 30), cv.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0))
# Print predicted classes.
samples/dnn/edge_detection.py
+3 -2
View File
@@ -99,10 +99,11 @@ def loadModel(args, engine):
return net
def apply_dexined(model, image):
t0 = cv.getTickCount()
out = model.forward()
t = (cv.getTickCount() - t0) / cv.getTickFrequency()
result,_ = post_processing(out, image.shape[:2])
t, _ = model.getPerfProfile()
label = 'Inference time: %.2f ms' % (t * 1000.0 / cv.getTickFrequency())
label = 'Inference time: %.2f ms' % (t * 1000.0)
cv.putText(image, label, (0, 15), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255))
cv.putText(result, label, (0, 15), cv.FONT_HERSHEY_SIMPLEX, 0.5, (255, 255, 255))
cv.imshow("Output", result)
samples/dnn/fast_neural_style.py
+3 -3
View File
@@ -34,14 +34,14 @@ while cv.waitKey(1) < 0:
swapRB=True, crop=False)
net.setInput(inp)
t0 = cv.getTickCount()
out = net.forward()
t = (cv.getTickCount() - t0) / cv.getTickFrequency()
out = out.reshape(3, out.shape[2], out.shape[3])
out = out.transpose(1, 2, 0)
t, _ = net.getPerfProfile()
freq = cv.getTickFrequency() / 1000
print(t / freq, 'ms')
print('%.2f ms' % (t * 1000.0))
if args.median_filter:
out = cv.medianBlur(out, args.median_filter)
samples/dnn/mask_rcnn.py
+3 -2
View File
@@ -105,7 +105,9 @@ while cv.waitKey(1) < 0:
# NOTE: In OpenCV 5.0, requesting 'detection_out_final' will fail if the .pbtxt
# does not register it as an output. See file header for details.
t0 = cv.getTickCount()
boxes, masks = net.forward(['detection_out_final', 'detection_masks'])
t = (cv.getTickCount() - t0) / cv.getTickFrequency()
numClasses = masks.shape[1]
numDetections = boxes.shape[2]
@@ -148,8 +150,7 @@ while cv.waitKey(1) < 0:
drawBox(*box)
# Put efficiency information.
t, _ = net.getPerfProfile()
label = 'Inference time: %.2f ms' % (t * 1000.0 / cv.getTickFrequency())
label = 'Inference time: %.2f ms' % (t * 1000.0)
cv.putText(frame, label, (0, 15), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0))
showLegend(classes)
samples/dnn/object_detection.cpp
+4 -4
View File
@@ -367,8 +367,11 @@ int main(int argc, char** argv)
}
preprocess(frame, net, Size(inpWidth, inpHeight));
TickMeter tickMeter;
vector<Mat> outs;
tickMeter.start();
net.forward(outs, net.getUnconnectedOutLayersNames());
tickMeter.stop();
classIds.clear();
confidences.clear();
@@ -378,13 +381,10 @@ int main(int argc, char** argv)
drawPred(classIds, confidences, boxes, frame, sans, stdSize, stdWeight, stdImgSize, stdThickness);
vector<double> layersTimes;
int imgWidth = max(frame.rows, frame.cols);
int size = static_cast<int>((stdSize * imgWidth) / (stdImgSize * 1.5));
int weight = static_cast<int>((stdWeight * imgWidth) / (stdImgSize * 1.5));
double freq = getTickFrequency() / 1000;
double t = net.getPerfProfile(layersTimes) / freq;
string label = format("FPS: %.2f", 1000/t);
string label = format("FPS: %.2f", 1000.0 / tickMeter.getTimeMilli());
putText(frame, label, Point(0, size), Scalar(0, 255, 0), sans, size, weight);
imshow(kWinName, frame);
}
samples/dnn/openpose.py
+3 -3
View File
@@ -82,7 +82,9 @@ while cv.waitKey(1) < 0:
inp = cv.dnn.blobFromImage(frame, inScale, (inWidth, inHeight),
(0, 0, 0), swapRB=False, crop=False)
net.setInput(inp)
t0 = cv.getTickCount()
out = net.forward()
t = (cv.getTickCount() - t0) / cv.getTickFrequency()
assert(len(BODY_PARTS) <= out.shape[1])
@@ -115,8 +117,6 @@ while cv.waitKey(1) < 0:
cv.ellipse(frame, points[idFrom], (3, 3), 0, 0, 360, (0, 0, 255), cv.FILLED)
cv.ellipse(frame, points[idTo], (3, 3), 0, 0, 360, (0, 0, 255), cv.FILLED)
t, _ = net.getPerfProfile()
freq = cv.getTickFrequency() / 1000
cv.putText(frame, '%.2fms' % (t / freq), (10, 20), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 0))
cv.putText(frame, '%.2f ms' % (t * 1000.0), (10, 20), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 0))
cv.imshow('OpenPose using OpenCV', frame)
samples/dnn/super_resolution.py
+3 -2
View File
@@ -132,12 +132,13 @@ def apply_super_resolution(net, image, args):
)
net.setInput(blob)
t0 = cv.getTickCount()
output = net.forward()
t = (cv.getTickCount() - t0) / cv.getTickFrequency()
result = postprocess_output(output, args, original_shape)
t, _ = net.getPerfProfile()
label = "Inference time: %.2f ms" % (t * 1000.0 / cv.getTickFrequency())
label = "Inference time: %.2f ms" % (t * 1000.0)
cv.putText(result, label, (10, 30), cv.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), 2)
return result