Abstract
The Radiographs are essential in clinical dentistry as they provide information that is invisible during an oral inspection. These images, however suffer from excess noise, low resolution, and poor contrast which impacts diagnosis accuracy. This study presents a two‑stage pipeline combining Enhanced Super‑Resolution GAN (ESRGAN) for radiograph enhancement followed by YOLOv8 for multi‑class dental anomaly detection. The proposed method involves the application of ESRGAN (Enhanced super-resolution generative adversarial network with adaptive dual perceptual loss) that improves image detail and sharpen resolution. A customized YOLOv8 object detection model which is trained to detect object and classify into six important dental conditions. The classification is Caries, Crown, Root Canal Treated (RCT), Restoration, Normal, and Badly Decayed teeth. The ESRGAN-enhanced images demonstrated high visual fidelity, achieving a Peak Signal-to-Noise Ratio (PSNR) of 28.7 dB and a Structural Similarity Index (SSIM) of 0.91. The proposed YOLOv8 model analyzes the images after being enhanced by ESRGAN. The YOLOv8 model was evaluated on 100 test images and achieved an overall mean Average Precision (mAP@0.5) of 56.9 % and mAP@0.5:0.95 of 41.6 %. The proposed model achieved Sensitivity (Recall) 0.942 and Specificity 0.919 for Crown detection. Detection of Caries and Badly Decayed teeth remained challenging, with lower sensitivity scores of 0.174 and 0.355, respectively. Specificity across classes ranged from 0.361 (RCT) to 0.887 (Caries), indicating variable false positive rates. The proposed pipeline demonstrated clinical potential by improving subtle structural visibility and supporting automated dental assessment. Future work will explore class‑specific augmentation and explainability tools to increase clinical utility. ESRGAN significantly improved the resolution and clarity of dental X-rays, enabling better visualization of fine details for accurate diagnosis. YOLOv8 effectively identified six dental conditions, achieving high accuracy for distinct classes like crowns and restorations.