MMRec

Why we included this project

MMRec is a research toolbox that gathers a large set of multimodal recommendation models into one codebase, so people studying how to recommend items from images, text, or other side data can compare approaches without reimplementing each paper. It includes reference implementations of well-known methods like VBPR, MMGCN, LATTICE, and BM3, plus newer graph-based and self-supervised models, all sharing the same data pipeline and evaluation setup. That shared infrastructure is the real draw: you can swap models, run them on identical datasets, and get comparable metrics instead of wrestling with divergent codebases. For researchers and engineers prototyping multimedia recommendation, it's a practical starting point for baselines and for seeing how different fusion and graph-learning strategies actually behave.