The first issue we encountered was model accuracy, which was initially about 5% to 20% but later increased to 90.31 percent. Second, we had challenges with Realtime Detection and Integration with Deployment. We have had troubles with Recommender's system as well.