arXiv reaDer
Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond
In this work, we introduce the Qwen-VL series, a set of large-scale vision-language models (LVLMs) designed to perceive and understand both texts and images. Starting from the Qwen-LM as a foundation, we endow it with visual capacity by the meticulously designed (i) visual receptor, (ii) input-output interface, (iii) 3-stage training pipeline, and (iv) multilingual multimodal cleaned corpus. Beyond the conventional image description and question-answering, we implement the grounding and text-reading ability of Qwen-VLs by aligning image-caption-box tuples. The resulting models, including Qwen-VL and Qwen-VL-Chat, set new records for generalist models under similar model scales on a broad range of visual-centric benchmarks (e.g., image captioning, question answering, visual grounding) and different settings (e.g., zero-shot, few-shot). Moreover, on real-world dialog benchmarks, our instruction-tuned Qwen-VL-Chat also demonstrates superiority compared to existing vision-language chatbots. Code, demo and models are available at https://github.com/QwenLM/Qwen-VL.
updated: Fri Oct 13 2023 02:41:28 GMT+0000 (UTC)
published: Thu Aug 24 2023 17:59:17 GMT+0000 (UTC)
参考文献 (このサイトで利用可能なもの) / References (only if available on this site)
被参照文献 (このサイトで利用可能なものを新しい順に) / Citations (only if available on this site, in order of most recent)
Amazon.co.jpアソシエイト