Abstract: In the digital turn of humanities, film studies has collided with new sparks after encountering digital tools and corpus thinking, which is particularly important for film history research. Using the image retrieval function of artificial intelligence tools, researchers can efficiently capture and summarize the required images and other information from the massive digital film corpus as a theoretical premise for more in-depth research. This article introduce the second round of retrieval experiments carried out in the early film corpus of the French National Audiovisual Research Center using the Snoop algorithm: Based on a review of the loop of motion image technology and theoretical reflection on how objects that appear repeatedly in film frames absorb, reflect and shape history, the study retrieves specific objects from the historical film corpus, proving that aesthetic associations will trigger historical analysis, thereby bringing the particularity of the background of the creation of the film corpus to the fore, and analyzing the potential and limitations of this research path.

Key words: digital humanities , film , studies , old film achive , image retrieval AI , metaphor of object