There are numerous techniques in film narratives to show the characters’ inner conflict and social background. This paper focuses on trying to show the characters’ psychology and social background by studying the spatial layout of housing in film narratives. Evolutionary deep learning methods are used to extract features of the spatial layout of housing in film and video images. Text analysis and other methods are used to obtain information about characters’ inner conflicts and social backgrounds and then encode them to transform textual information into quantifiable data. The Pearson correlation coefficients between the housing space layout and the coded character’s inner conflict and social background are calculated to analyze the correlation between the housing space layout and the character’s inner conflict and social background in the film narrative. This paper’s model, which extracts features from the housing spatial layout, achieves mIoU, MPA, and PE evaluation indexes of 80.57%, 89.93%, and 7.62%, respectively, outperforming other models. The model presented in this paper performs the best in PE evaluation metrics, reaching a pixel-level error rate of 6.26% on the LSNU test set. By calculating the Pearson correlation coefficient between housing space layout and characters’ inner conflict and social background, this paper fully demonstrates the reasonableness of the use of housing space layout in film narratives to visually display characters’ inner conflict and social background, as well as the ingenuity of the technique.