AdaptLLM/remote-sensing-Qwen2.5-VL-3B-Instruct
AdaptLLM/remote-sensing-Qwen2.5-VL-3B-Instruct is a 3 billion parameter multimodal large language model (MLLM) developed by AdaptLLM, fine-tuned from Qwen2.5-VL-3B-Instruct. This model is specifically adapted for remote sensing applications through post-training on a dedicated dataset. It excels at visual question answering and descriptive tasks within the remote sensing domain, offering specialized capabilities for analyzing satellite and aerial imagery.
Loading preview...
Overview
AdaptLLM/remote-sensing-Qwen2.5-VL-3B-Instruct is a 3 billion parameter multimodal large language model (MLLM) specifically adapted for remote sensing tasks. Developed by AdaptLLM, this model is a post-trained version of Qwen2.5-VL-3B-Instruct, leveraging domain-adaptive techniques to enhance its performance on specialized visual data. It was trained for one epoch with a learning rate of 1e-5 and a global batch size of 128, using the remote-sensing-visual-instructions dataset.
Key Capabilities
- Domain-Adapted Multimodal Understanding: Specialized in processing and interpreting remote sensing imagery combined with textual queries.
- Visual Question Answering (VQA): Designed to answer questions based on remote sensing images, providing relevant descriptions and insights.
- Compatibility: Built upon the Qwen2.5-VL-3B-Instruct architecture, ensuring compatibility with existing Qwen2.5-VL tooling and workflows.
Good For
- Remote Sensing Analysis: Ideal for applications requiring detailed analysis and interpretation of satellite and aerial imagery.
- Geospatial Intelligence: Useful for tasks involving understanding geographical features, land use, and environmental changes from visual data.
- Research in Domain Adaptation: Provides a strong baseline for further research into adapting MLLMs to specific scientific or industrial domains.