小米影像大脑升级为 AISP，打造首个 AI 大模型计算摄影平台
北京时间 2 月 27 日晚间，小米在小米 14 Ultra 新品发布会上宣布，小米影像大脑全面升级为“首个 AI 大模型计算摄影平台”，定名为“Xiaomi AISP”。
据介绍，小米 AISP 通过小米澎湃 OS 深入硬件层，全面整合 CPU、GPU、NPU 和 ISP 的算力，可达 60TOPS。这标志着小米在计算摄影领域取得了重大突破，为用户带来更出色的影像体验。
小米 AISP 拥有三大核心技术优势：
* 全链路 AI 大模型：小米 AISP 采用全链路 AI 大模型，涵盖图像处理、视频处理、显示增强等多个环节。通过端到端的训练，AI 大模型可以学习图像和视频的底层规律，实现更准确、更自然的图像和视频处理效果。
* 异构计算架构：小米 AISP 采用异构计算架构，充分利用 CPU、GPU、NPU 和 ISP 的不同优势。通过合理的算力分配，小米 AISP 可以实现高性能、低功耗的影像处理。
* 自研算法与模型：小米 AISP 搭载了小米自研的算法和模型，针对不同场景和用户需求进行了深度优化。这些算法和模型可以有效提升图像和视频的质量，满足用户对高品质影像的需求。
小米 AISP 的应用场景广泛，包括：
* 夜景拍摄：小米 AISP 可以大幅提升夜景拍摄的亮度、清晰度和色彩还原度，让用户轻松拍出清晰明亮的夜景照片。
* 人像拍摄：小米 AISP 可以精准识别人物主体，并针对性地优化肤色、五官和背景虚化效果，打造自然美观的人像照片。
* 视频拍摄：小米 AISP 可以实时优化视频画质，提升视频的稳定性、清晰度和色彩表现力，让用户轻松记录精彩瞬间。
小米 AISP 的推出，标志着小米在计算摄影领域迈出了重要一步。通过 AI 大模型、异构计算架构和自研算法的结合，小米 AISP 为用户带来了更出色的影像体验，进一步提升了小米手机的影像实力。
英语如下：
**Headline: Xiaomi’s Image Brain Evolves: AI Large Model Ushersin New Era of Computational Photography**
**Keywords:** AI photography, large model, enhanced computing power
**News Content:** Xiaomi Image Brain Upgrades to AISP, Creating the First AI Large Model Computational Photography Platform
On the evening of February 27, Beijing time, Xiaomi announced at the Xiaomi 14 Ultra new product launch that Xiaomi Image Brain has been fully upgraded to the “first AI large model computational photography platform,” named “Xiaomi AISP.”
According to theintroduction, Xiaomi AISP integrates the computing power of CPU, GPU, NPU, and ISP through Xiaomi’s Pengpai OS, deeply into the hardware layer, reaching 60TOPS. This marks a major breakthrough for Xiaomi in the field of computational photography, bringing users a more outstanding imaging experience.
Xiaomi AISP has three core technical advantages:
* Full-chain AI large model: Xiaomi AISP adopts a full-chain AI large model, covering multiple links such as image processing, video processing, and display enhancement. Through end-to-end training, the AI large model can learn the underlying laws of images andvideos, achieving more accurate and natural image and video processing effects.
* Heterogeneous computing architecture: Xiaomi AISP adopts a heterogeneous computing architecture, making full use of the different advantages of CPU, GPU, NPU, and ISP. Through reasonable computing power allocation, Xiaomi AISP can achieve high-performance, low-power image processing.
* Self-developed algorithms and models: Xiaomi AISP is equipped with Xiaomi’s self-developed algorithms and models, which have been deeply optimized for different scenarios and user needs. These algorithms and models can effectively improve the quality of images and videos, meeting users’ demands for high-quality images.
Xiaomi AISP has a wide range of application scenarios, including:
* Night photography: Xiaomi AISP can significantly improve the brightness, clarity, and color reproduction of night shots, allowing users to easily take clear and bright night photos.
* Portrait photography: Xiaomi AISP can accurately identify human subjects and specifically optimize skin tone, facial features, and background blur effects, creating natural and beautiful portraits.
* Video shooting: Xiaomi AISP can optimize video quality in real time, enhancing the stability, clarity, and color performance of videos, allowing users to easily record wonderful moments.
The launch of Xiaomi AISP marks an importantstep for Xiaomi in the field of computational photography. Through the combination of AI large models, heterogeneous computing architecture, and self-developed algorithms, Xiaomi AISP brings users a more outstanding imaging experience, further enhancing the imaging capabilities of Xiaomi phones.
【来源】https://www.ithome.com/0/751/474.htm
Views: 0