Skip to content

CUDA: Handle fusion of conv+eltwise in case of multi-output node (i.e. Split)#27326

Merged
asmorkalov merged 1 commit intoopencv:4.xfrom
dkurt:handle_multi_output_eltwise_fusion
May 19, 2025
Merged

CUDA: Handle fusion of conv+eltwise in case of multi-output node (i.e. Split)#27326
asmorkalov merged 1 commit intoopencv:4.xfrom
dkurt:handle_multi_output_eltwise_fusion

Conversation

@dkurt
Copy link
Copy Markdown
Member

@dkurt dkurt commented May 17, 2025 •
edited
Loading

Pull Request Readiness Checklist

Enables YOLO11n with CUDA backend

resolves #26566

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

  • I agree to contribute to the project under Apache 2 License.
  • 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
  • The PR is proposed to the proper branch
  • There is a reference to the original bug report and related work
  • There is accuracy test, performance test and test data in opencv_extra repository, if applicable
    Patch to opencv_extra has the same branch name.
  • The feature is well documented and sample code can be built with the project CMake

@dkurt dkurt changed the title Handle fusion of conv+eltwise in case of multi-output node (i.e. Split) CUDA: Handle fusion of conv+eltwise in case of multi-output node (i.e. Split) May 17, 2025
@dkurt
Copy link
Copy Markdown
Member Author

dkurt commented May 17, 2025 •
edited
Loading 

import numpy as np
import cv2 as cv

net = cv.dnn.readNet("yolo11n.onnx")

inp = np.random.rand(1, 3, 640, 640).astype(np.float32)
net.setInput(inp)
net.setPreferableTarget(cv.dnn.DNN_TARGET_CPU)
ref = net.forward()

net = cv.dnn.readNet("yolo11n.onnx")

net.setInput(inp)
net.setPreferableBackend(cv.dnn.DNN_BACKEND_CUDA)
net.setPreferableTarget(cv.dnn.DNN_TARGET_CUDA)
out = net.forward()

print("ref shape", ref.shape)
print("out shape", out.shape)
print("diff", np.max(np.abs(ref - out)))
ref shape (1, 84, 8400)
out shape (1, 84, 8400)
diff 0.0029296875

@dkurt dkurt added this to the 4.12.0 milestone May 17, 2025
@dkurt dkurt requested a review from asmorkalov May 17, 2025 11:08
@asmorkalov asmorkalov self-assigned this May 19, 2025
@asmorkalov asmorkalov merged commit 9d2d927 into opencv:4.x May 19, 2025
103 of 109 checks passed
@dkurt dkurt deleted the handle_multi_output_eltwise_fusion branch May 19, 2025 07:48
@asmorkalov asmorkalov mentioned this pull request May 27, 2025
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@asmorkalov asmorkalov asmorkalov approved these changes

Assignees

@asmorkalov asmorkalov

Labels

bug category: dnn

Projects

None yet

Milestone

4.12.0

Development

Successfully merging this pull request may close these issues.

YOLO11 models do not operate with the CUDA and CUDA FP16 targets (OpenCV 4.10.0)

2 participants

@dkurt @asmorkalov