Segfault when Net::Impl::forward(OutputArrayOfArrays outputBlobs, const std::vector<String>& outBlobNames) is called with empty outBlobNames #27352
Description
System Information
- Ubuntu 24.04 x64
- gcc 13.3.0
- OpenCV 4.11.0
Detailed description
Calling the forward() method with empty outBlobNames causes a segfault with no helpful message to debug. The method works fine if correct output layer names are provided using net.getUnconnectedOutLayersNames();
[1] 11096 segmentation fault (core dumped) ./bug
using GDB i could backtrace it to
#0 0x00007f6df3876ec9 in cv::dnn::dnn4_v20241223::Net::Impl::getLatestLayerPin(std::vector<cv::dnn::dnn4_v20241223::detail::LayerPin, std::allocator<cv::dnn::dnn4_v20241223::detail::LayerPin> > const&) const () from /opt/opencv-4.11.0/lib/libopencv_dnn.so.411
No symbol table info available.
#1 0x00007f6df388daf9 in cv::dnn::dnn4_v20241223::Net::Impl::forward(cv::_OutputArray const&, std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > > const&) () from /opt/opencv-4.11.0/lib/libopencv_dnn.so.411
No symbol table info available.
#2 0x00007f6df386dfed in cv::dnn::dnn4_v20241223::Net::forward(cv::_OutputArray const&, std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > > const&) () from /opt/opencv-4.11.0/lib/libopencv_dnn.so.411
No symbol table info available.
where getLatestLayerPin() has no check for empty, which results in de-referencing an invalid iterator.
LayerPin Net::Impl::getLatestLayerPin(const std::vector<LayerPin>& pins) const
{
return *std::max_element(pins.begin(), pins.end());
}Steps to reproduce
Compile and run the following sample
#include <opencv2/dnn.hpp>
#include <opencv2/imgcodecs.hpp>
#include <iostream>
int main()
{
try
{
// load some model
cv::dnn::Net net = cv::dnn::readNetFromONNX("yolo_nas_s.onnx");
cv::Size network_input_size(640, 640);
cv::Mat image(network_input_size, CV_8UC3);
cv::randu(image, cv::Scalar(0, 0, 0), cv::Scalar(255, 255, 255));
// backend agnostic
net.setPreferableBackend(cv::dnn::DNN_BACKEND_OPENCV);
net.setPreferableTarget(cv::dnn::DNN_TARGET_CPU);
cv::Mat blob = cv::dnn::blobFromImage(
image, 1, network_input_size, cv::Scalar(0, 0, 0), true, false, CV_32F);
net.setInput(blob);
std::vector<cv::Mat> outs;
std::vector<cv::String> outNames; // Empty vector
// ideally if i do the following, it works
// outNames = net.getUnconnectedOutLayersNames();
net.forward(outs, outNames); // <-- Segfaults here
}
catch (const cv::Exception &e)
{
// should throw this atleast to know what caused the error
std::cerr << "OpenCV Exception: " << e.what() << std::endl;
return 1;
}
catch (const std::exception &e)
{
std::cerr << "Standard Exception: " << e.what() << std::endl;
return 1;
}
return 0;
}NOTE: Uploading the onnx model is failing here, likely due to file size, ive tried zip version too
Issue submission checklist
- I report the issue, it's not a question
- I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution
- I updated to the latest OpenCV version and the issue is still there
- There is reproducer code and related data files (videos, images, onnx, etc)