Scribd Logo
Upload

Welcome to Scribd!

  • 7 views

    Detect

    Uploaded by

    Shumail Nazir
    This C++ program loads a TorchScript model and uses it to perform object detection on video frames from a video file. It preprocesses each frame, runs detection on the model, extracts the bounding boxes and class labels, draws these on the frame, and displays the annotated video frame by frame. It loads the model, opens the video, processes each frame by running detection and drawing results, and finally releases resources before ending.

    Copyright:

    © All Rights Reserved
    Download as TXT, PDF, TXT or read online from Scribd

    Detect

    Uploaded by

    Shumail Nazir
    7 views2 pages
    This C++ program loads a TorchScript model and uses it to perform object detection on video frames from a video file. It preprocesses each frame, runs detection on the model, extracts the bounding boxes and class labels, draws these on the frame, and displays the annotated video frame by frame. It loads the model, opens the video, processes each frame by running detection and drawing results, and finally releases resources before ending.

    Copyright

    © © All Rights Reserved

    Available Formats

    TXT, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This C++ program loads a TorchScript model and uses it to perform object detection on video frames from a video file. It preprocesses each frame, runs detection on the model, extracts the bounding boxes and class labels, draws these on the frame, and displays the annotated video frame by frame. It loads the model, opens the video, processes each frame by running detection and drawing results, and finally releases resources before ending.

    Copyright:

    © All Rights Reserved
    Download as TXT, PDF, TXT or read online from Scribd
    Download as txt, pdf, or txt
    7 views2 pages

    Detect

    Uploaded by

    Shumail Nazir
    This C++ program loads a TorchScript model and uses it to perform object detection on video frames from a video file. It preprocesses each frame, runs detection on the model, extracts the bounding boxes and class labels, draws these on the frame, and displays the annotated video frame by frame. It loads the model, opens the video, processes each frame by running detection and drawing results, and finally releases resources before ending.

    Copyright:

    © All Rights Reserved
    Download as TXT, PDF, TXT or read online from Scribd
    Download as txt, pdf, or txt
    You are on page 1of 2

    #include <torch/script.

    h>
    #include <opencv2/opencv.hpp> // Include the OpenCV header

    int main() {
    std::string model_path = "/home/ubuntu/exported_model.pt";
    std::string video_path = "/home/ubuntu/Downloads/remmie.mp4";

    try {
    // Load the TorchScript model
    torch::jit::script::Module model;
    model = torch::jit::load(model_path);
    std::cout << "Model is Loaded Successfully" << std::endl;
    model.eval();

    // Open the video file


    cv::VideoCapture video(video_path);
    if (!video.isOpened()) {
    std::cerr << "Error opening the video file" << std::endl;
    return -1;
    }

    // Process each frame in the video


    cv::Mat frame;
    while (video.read(frame)) {
    // Convert the frame to a Torch tensor and preprocess if needed
    cv::Mat input;
    cv::cvtColor(frame, input, cv::COLOR_BGR2RGB); // Convert BGR to RGB
    cv::resize(input, input, cv::Size(224, 224)); // Resize to model's
    input size
    input.convertTo(input, CV_32FC3, 1.0 / 255); // Normalize pixel
    values

    // Create a Torch tensor from the preprocessed input


    torch::Tensor inputTensor = torch::from_blob(
    input.data, {1, input.rows, input.cols,
    input.channels()}).permute({0, 3, 1, 2});

    // Perform object detection on the frame


    std::vector<torch::jit::IValue> inputs;
    inputs.emplace_back(inputTensor);
    at::Tensor outputTensor = model.forward(inputs).toTensor();

    // Extract bounding box coordinates and class labels from the detection
    results
    auto output = outputTensor.accessor<float, 2>();
    std::vector<cv::Rect> boxes;
    std::vector<int> labels;
    for (int i = 0; i < output.size(0); i++) {
    float label = output[i][0];
    float xmin = output[i][1] * frame.cols;
    float ymin = output[i][2] * frame.rows;
    float xmax = output[i][3] * frame.cols;
    float ymax = output[i][4] * frame.rows;

    boxes.emplace_back(cv::Rect(cv::Point(xmin, ymin), cv::Point(xmax,


    ymax)));
    labels.emplace_back(static_cast<int>(label));
    }
    // Draw rectangles and labels on the frame
    for (size_t i = 0; i < boxes.size(); i++) {
    cv::rectangle(frame, boxes[i], cv::Scalar(0, 255, 0), 2);
    cv::putText(frame, std::to_string(labels[i]), cv::Point(boxes[i].x,
    boxes[i].y - 5),
    cv::FONT_HERSHEY_SIMPLEX, 0.8, cv::Scalar(0, 255, 0),
    2);
    }

    // Display or save the modified frame


    cv::imshow("Video", frame);
    cv::waitKey(1); // Adjust the delay between frames as needed

    // Break the loop if the 'q' key is pressed


    if (cv::waitKey(1) == 'q') {
    break;
    }
    }

    // Release the video file and clean up


    video.release();
    cv::destroyAllWindows();

    } catch (const c10::Error& e) {


    std::cerr << "Error loading the model: " << e.msg() << std::endl;
    return -1;
    }

    return 0;
    }

    You might also like