4 个月前

BT-Adapter:无需视频指令调优即可实现视频对话

BT-Adapter:无需视频指令调优即可实现视频对话

摘要

近期在大规模语言模型(Large Language Models, LLM)方面的进展推动了图像-语言对话代理的多种进步,但如何构建一个高效的基于视频的对话系统仍处于探索阶段。考虑到大规模语言模型和视觉主干网络的庞大体量,留给有效时序建模的GPU内存非常有限,而时序建模对于理解和回应视频内容至关重要。为此,我们提出了一种新的方法——分支时序适配器(Branching Temporal Adapter, BT-Adapter),用于将图像-语言预训练模型扩展到视频领域。具体而言,BT-Adapter作为预训练视觉编码器旁的一个即插即用的时序建模分支,在保持主干网络冻结的情况下进行微调。只需一次预训练,BT-Adapter即可无缝集成到所有使用此版本CLIP的图像对话模型中,实现无需视频指令的视频对话功能。此外,我们在分支内部开发了一种独特的非对称标记掩码策略,并为BT-Adapter设计了定制化的训练任务,从而加速收敛并获得更好的结果。得益于BT-Adapter,我们能够在不增加过多GPU成本的情况下增强现有多模态对话模型的视频理解能力。无需额外复杂的配置,BT-Adapter实现了以下几点:(1) 在各种视频任务上以较少的GPU小时数达到了最先进的零样本性能;(2) 无需任何视频指令微调的情况下优于当前的视频聊天机器人;(3) 经过视频指令微调后,在视频聊天方面取得了远超以往最佳水平的结果。

代码仓库

farewellthree/BT-Adapter
官方
pytorch
GitHub 中提及

基准测试

基准方法指标
vcgbench-diverse-on-videoinstructBT-Adapter
Consistency: 2.27
Contextual Understanding: 2.59
Correctness of Information: 2.20
Dense Captioning: 1.03
Detail Orientation: 2.62
Reasoning: 3.62
Spatial Understanding: 2.35
Temporal Understanding: 1.29
mean: 2.19
video-based-generative-performanceBT-Adapter (zero-shot)
Consistency: 2.2
Contextual Understanding: 2.89
Correctness of Information: 2.16
Detail Orientation: 2.46
Temporal Understanding: 2.13
mean: 2.46
video-based-generative-performanceBT-Adapter
Consistency: 2.46
Contextual Understanding: 3.27
Correctness of Information: 2.68
Detail Orientation: 2.69
Temporal Understanding: 2.34
mean: 2.69
video-based-generative-performance-1BT-Adapter
gpt-score: 2.68
video-based-generative-performance-1BT-Adapter (zero-shot)
gpt-score: 2.16
video-based-generative-performance-2BT-Adapter
gpt-score: 2.46
video-based-generative-performance-2BT-Adapter (zero-shot)
gpt-score: 2.2
video-based-generative-performance-3BT-Adapter
gpt-score: 3.27
video-based-generative-performance-3BT-Adapter (zero-shot)
gpt-score: 2.89
video-based-generative-performance-4BT-Adapter (zero-shot)
gpt-score: 2.46
video-based-generative-performance-4BT-Adapter
gpt-score: 2.69
video-based-generative-performance-5BT-Adapter (zero-shot)
gpt-score: 2.13
video-based-generative-performance-5BT-Adapter
gpt-score: 2.34
video-question-answering-on-activitynet-qaBT-Adapter (zero-shot)
Accuracy: 46.1
Confidence score: 3.6
zero-shot-video-retrieval-on-activitynetBT-Adapter
text-to-video R@1: 37.0
text-to-video R@10: 78.9
text-to-video R@5: 66.7
zero-shot-video-retrieval-on-didemoBT-Adapter
text-to-video R@1: 35.6
text-to-video R@10: 72.6
text-to-video R@5: 61.9
zero-shot-video-retrieval-on-lsmdcBT-Adapter
text-to-video R@1: 19.5
text-to-video R@10: 45.0
text-to-video R@5: 35.9
zero-shot-video-retrieval-on-msr-vttBT-Adapter
text-to-video R@1: 40.9
text-to-video R@10: 73.5
text-to-video R@5: 64.7
zeroshot-video-question-answer-on-activitynetBT-Adapter (zero-shot)
Accuracy: 46.1
Confidence Score: 3.2
zeroshot-video-question-answer-on-msrvtt-qaBT-Adapter (zero-shot)
Accuracy: 51.2
Confidence Score: 2.9
zeroshot-video-question-answer-on-msrvtt-qaBT-Adapter (zero-shot)
Accuracy: 51.2
Confidence Score: 2.9
zeroshot-video-question-answer-on-msvd-qaBT-Adapter (zero-shot)
Accuracy: 67.0
Confidence Score: 3.6
zeroshot-video-question-answer-on-msvd-qaBT-Adapter (zero-shot)
Accuracy: 67.0
Confidence Score: 3.6

用 AI 构建 AI

从想法到上线——通过免费 AI 协同编程、开箱即用的环境和市场最优价格的 GPU 加速您的 AI 开发

AI 协同编程
即用型 GPU
最优价格
立即开始

Hyper Newsletters

订阅我们的最新资讯
我们会在北京时间 每周一的上午九点 向您的邮箱投递本周内的最新更新
邮件发送服务由 MailChimp 提供
BT-Adapter:无需视频指令调优即可实现视频对话 | 论文 | HyperAI超神经