Few of the biggest challenges in Image Segmentation are manual marking of objects and the segmentation accuracy. Due to multiple manual iterations, performance degrades, and real time segmentation can’t be achieved in digital images or videos. Best way to achieve automation is to use Deep Learning based ML techniques like RMASK-CNN, YOLO etc. Accuracy is another challenge where we have to deal with the False Positives and False Negatives. Segmentation process must be very accurate with near to zero FPs and FNs especially if it is going to get used in medical field or in security field. Pre-processing is necessary to improve the quality of the supplied image or video to achieve high accuracy during segmentation. This research aims to develop a framework that will enable automated and accurate segmentation in digital image also gives clear guidelines on configurations and process flows which can help users to customize and use it in any field. Automation using deep learning brings another challenge i.e. to maintain big datasets and keep enhancing it to achieve more accuracy. To deal with heavy datasets we have proposed to use ensembling techniques. Also, it should be based on open-source technology to have control on overall cost. At the end clear roadmap and limitations to be declared so that researchers get clear directions on future work.