In this paper, we introduce a new method for segmenting text lines from scanned handwritten images in the Marwari script. Marwari is mainly spoken in the Indian state of Rajasthan and nearby areas. Handwritten Marwari documents are challenging to digitize due to different handwriting styles and complex connections between letters. Preserving and digitizing these documents is important for understanding Marwari history, but not much work has been done in this area. Deep learning techniques can help detect text and speed up the digitization process. Our approach uses image processing, deep learning, and line segmentation to accurately identify text lines in handwritten Marwari documents. Tests on a dataset of scanned Marwari documents show that our method works well and is more effective than other existing techniques. The main contribution of our work is successfully segmenting very long and interconnected words, which will be helpful for further research.