[AAAI 2025]HiCM²: Hierarchical Compact Memory Modeling for Dense Video Captioning

Official implementation of "LLaVA Needs More Knowledge: Retrieval-Augmented NLG with Knowledge Graph for Explaining Thoracic Pathologies" (AAAI 2025). This repo provides the KG-LLaVA framework, integrating KG-based Retrieval-Augmented Generation (KG-RAG) with Vision-Language Models to generate NLEs for medical imaging.

[CVPR 2024] Do you remember? Dense Video Captioning with Cross-Modal Memory Retrieval

This is the official source code for CVPR 2024 paper [WWW: A Unified Framework for Explaining What, Where and Why of Neural Networks by Interpretation of Neuron Concepts]