In this study, we propose a new method for image recognition based on computer vision, combining deep convolutional neural networks (DCNN), attention mechanisms, and generative adversarial networks (GAN) to enhance the accuracy of object identification. First, DCNN was utilized to extract features from images to automatically identify objects within them. To better capture image features at different scales, a multi-scale attention mechanism was introduced, which significantly improves the accuracy of feature extraction by focusing on key areas in the image. Additionally, GAN-generated image data was used to augment the training dataset, thereby enhancing the model’s generalization capability when dealing with diverse real-world scenarios. Experimental results demonstrate that the proposed algorithm achieves high precision across multiple public datasets, particularly excelling in complex scenes by notably reducing false and missed detections. Compared with traditional methods, this approach improves recognition accuracy by approximately $9 \%$, providing robust technical support for rapid and precise image recognition tasks.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Application of computer vision algorithm based on deep learning in image recognition


    Beteiligte:
    Yang, Zixuan (Autor:in)


    Erscheinungsdatum :

    23.10.2024


    Format / Umfang :

    687007 byte




    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch




    Editorial- Deep Learning for Computer Vision

    Girshick, Ross / Kokkinos, Iasonas / Laptev, Ivan et al. | British Library Online Contents | 2017


    Deep Learning Based Self Driving Cars Using Computer Vision

    Bhaggiaraj, S. / Priyadharsini, M. / Karuppasamy, K. et al. | IEEE | 2023


    A Modified YOLOv4 Deep Learning Network for Vision-Based UAV Recognition

    Farzaneh Dadrass Javan / Farhad Samadzadegan / Mehrnaz Gholamshahi et al. | DOAJ | 2022

    Freier Zugriff