<p>Text detection in natural acts is one of the vital jobs in computer vision applications. Text detection helps to understand the scene by providing valuable information about various aspects such as signboard, product label, navigation etc. In this context, recognition and detection of text present in the scenic images has gained significant attention. However, the text detection process poses unique challenges due to irregular text layouts and complex backgrounds. In addition, other factors such as irregular shapes, poor illumination, and varying text size increases the difficulty of text detection. Existing datasets used in the text detection process consist of structured data which is not suitable to detect curved data or small sized texts. Text recognition models trained on structured data often fail to detect the texts accurately. This research created a dataset consisting of curved text with varying sizes to maximize the accuracy of detection. An L-YoloV8 (Large You Look Only Once) model is employed in this work for recognizing curved text in natural scenic images. The model is trained to detect texts from both planar and non-planar surfaces including empty text detection. Outputs illustration that the L-YoloV8 approach achieves a phenomenal accuracy of 91.85% for text detection compared to other techniques.</p>