RSGPT A Remote Sensing Vision Language Model and Benchmark

• 论文名称: RSGPT: A Remote Sensing Vision Language Model and Benchmark
• 模型架构: MLLM
• Visual Encoder: Transformer
• Text Encoder: Transformer
• Model Details: Vision Encoder：EVAText Encoder：Vicuna
• Task: Image Caption, RS VQA
• Link: https://arxiv.org/abs/2307.15266
• Code/Project: -
• Short Summary: 1. 提出RSICap数据集，基于 DOTA 目标检测数据集构建了 RSICap2. 提出RSIEval评估集3. RSGPT模型，现成的冻结的预训练图像编码器（EVA-G）和大型语言模型（vicuna7b，vicuna13b）构成了该模型的基础，并通过微调Q-Former和线性层结构
• Published in: Arxiv 2023