Deep Learning-Based 3D Indoor Room Layout Prediction with ConvBi-LSTM Models

Estimating room arrangement has become more and more significant in recent years due to its many uses. This method is very challenging because of the many factors that affect the room image, such as clutter, occlusions, illuminations, etc. It is essential to accurately determine the 3D layout of the area from a single 2D room image. The available approaches were mostly focused on determining the 3D arrangement and lacked many features. Many features must be added to a model in order for it to generate good predictions. To this degree, a strong foundation for estimating 3D layouts in interior environments was introduced by the suggested model. The input image is first pre-processed before being placed via layout estimation, where our suggested model forecasts the image's edge maps and semantic labels. The suggested framework used the Deep ConvBi-LSTM model for prediction, and the remora optimization algorithm (ROA) developed and maximized a score function to extract the best 2D layout from the candidate set. Lastly, using the layout coordinates and camera orientations, the 3D output is recreated from the 2D layout. The suggested model's effectiveness in delivering the intended performance was demonstrated by the experimental findings.