MUMBAI, India, March 13 -- Intellectual Property India has published a patent application (202621011111 A) filed by Symbiosis International, Pune, Maharashtra, on Feb. 2, for 'ai-generated image detection system using vision transformers and convolutional neural network architectures.'
Inventor(s) include Dr. Shreyas Rajendra Hole.
The application for the patent was published on March 13, under issue no. 11/2026.
According to the abstract released by the Intellectual Property India: "The present invention discloses an AI-generated image detection system (100) for classifying digital images as real or AI-generated. The system comprises an image input module (110), a preprocessing module (120) with resizing and normalization units, a feature extraction module (130) incorporating a Vision Transformer network (131), VGG16 (132), and AlexNet (133) architectures, a classification module (140) with global average pooling and sigmoid activation, an output module (150), and a training module (160). The Vision Transformer network utilizes self-attention mechanisms to capture global contextual dependencies by processing images as sequences of patches with positional encodings. The system achieves accuracy exceeding 98% with AUC of 0.992 through transfer learning from pretrained models and GPU-accelerated training. The invention addresses limitations of conventional deepfake detection methods by combining transformer-based global perception with CNN-based local feature extraction for robust identification of subtle textural inconsistencies in synthetic content generated by modern diffusion models and GANs."
Disclaimer: Curated by HT Syndication.
