Image segmentation plays an important role in computer vision and is a key element in machine learning. Image segmentation is a critical stage in many computer vision applications that helps provide accurate and relevant data by improving object localization, recognition, feature extraction, and classification. Entire comparative study show case two experiments.First consists of the traditional approach of passing the images directly into an image segmentation model. In second approach tried with a variety of images with the use of a hybrid model. First model is the image segmentation model with DeepLabV3 architecture. The second experiment is a strategy to enhance image segmentation and achieve better results. In the second experiment, the output of the object detection model, YOLOv5, serves as the input to the image segmentation model. The person segmentation dataset was used to train the image segmentation model. With use of a distinctive dataset to assess the performance of these surveys in the end. Our experiments show that the second strategy performs better than the first approach in comparison.