--- title: Vision and Sensing | Privacy-Centered AI Pipelines description: Explore hardware-aware algorithms and secure sensing pipelines. Sony AI redefines imaging systems through privacy-centered architecture and AI research. --- # **Vision and Sensing** Our research aims to revolutionize how vision AI is deployed and applied in the real world, by developing compact, high-performance foundation models and robust imaging pipelines. [Explore Vision & Sensing Projects ](https://ai.sony/research-areas/vision-sensing#vision-and-sensing-projects) ## **Reimagining Vision AI** Building compact foundation models and robust pipelines for vision and sensing. Sony AI is exploring cutting-edge vision technologies to maximize vision and sensing capabilities using AI. This includes a compact and powerful vision foundation model series for vision, model compression, and optimization that will help shape the future of edge AI. At the same time, it creates an optimal AI pipeline design for both imaging and sensing to leverage the latest advancements in computer vision and machine learning methods. ![A man in glasses works on a Vaio laptop in an office, illuminated by pink light](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Images/A%20man%20in%20glasses%20works%20on%20a%20Vaio%20laptop%20in%20an%20office%2c%20illuminated%20by%20pink%20light.webp?width=1200&height=900&name=A%20man%20in%20glasses%20works%20on%20a%20Vaio%20laptop%20in%20an%20office%2c%20illuminated%20by%20pink%20light.webp) ## **Breakthroughs in Vision & Sensing** Learn about our global initiatives unlocking complex application environments and new business opportunities. See how these projects push the boundaries of vision and sensing. OUR WORK ### Argus: A Compact and Versatile Vision Foundation Model Argus is a powerful vision foundation model designed to handle 12 different computer vision tasks—from smart cities to retail—on both cloud and edge devices. By training on just a fraction of the usual data, we’ve created a high-performance backbone that is versatile, practical, and significantly more accessible to deploy. [Learn More ](https://cvpr.thecvf.com/media/PosterPDFs/CVPR%202025/32530.png?t=1748361487.889834) OUR WORK ### Stretching Each Dollar: Diffusion Training from Scratch on a Micro-Budget We’ve developed a strategy to democratize high-end generative AI by training a 1.1 billion parameter model for under $2,000—over 100x less than traditional stable diffusion costs. Using smart masking and synthetic data, we achieve high-quality image generation while proving that top-tier research doesn't always require massive computational budgets. [Learn More ](https://ai.sony/hubfs/Stretching%20Each%20Dollar-%20Diffusion%20Training%20from%20Scratch%20on%20a%20Micro-Budget.pdf?hsLang=en) OUR WORK ### Beyond RGB: Adaptive Parallel Processing for RAW Object Detection Traditional image processing is built for human eyes, but it often throws away data that AI needs to see clearly. Our Raw Adaptation Module (RAM) mimics the human visual system’s parallel processing to extract more detail directly from raw sensor data, leading to state-of-the-art object detection even in difficult lighting and high-dynamic-range environments. [Learn More ](https://arxiv.org/abs/2503.13163) ## **Deep Dive** Into Our Research Discover how we are leveraging breakthroughs in computer vision and machine learning to unlock new business opportunities and navigate complex application environments. ![Futuristic server room with a glowing dome featuring circuit patterns and a brain icon](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Images/Futuristic%20server%20room%20with%20a%20glowing%20dome%20featuring%20circuit%20patterns%20and%20a%20brain%20icon.webp?width=1200&height=900&name=Futuristic%20server%20room%20with%20a%20glowing%20dome%20featuring%20circuit%20patterns%20and%20a%20brain%20icon.webp) AI Pipeline Our goal is to maximize imaging and sensing capabilities using AI. Traditional RGB pipelines are optimized for human perception, and often discard information that could be critical for specific applications, such as scientific analysis or autonomous navigation. We are creating AI-optimized pipelines for computer vision.  ![A person types on a laptop, with digital gears and flow lines emerging](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Images/A%20person%20types%20on%20a%20laptop%2c%20with%20digital%20gears%20and%20flow%20lines%20emerging.webp?width=1200&height=900&name=A%20person%20types%20on%20a%20laptop%2c%20with%20digital%20gears%20and%20flow%20lines%20emerging.webp) Vision Foundation Models Our compact vision foundation models will help shape the future of edge AI, especially given the limited memory and computational capabilities of edge devices. These models provide numerous advantages beyond just deployment flexibility. They require fewer compute resources, enable faster inference speeds, and contribute to smaller carbon footprints. ## **Awards** & Recognitions ![Neural Information Processing Systems logo](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Logos/logo-NIPS.png?width=192&height=86&name=logo-NIPS.png) ![Neural Information Processing Systems logo](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Logos/logo-NIPS.png?width=192&height=86&name=logo-NIPS.png) ![Neural Information Processing Systems logo](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Logos/logo-NIPS.png?width=192&height=86&name=logo-NIPS.png) ![Neural Information Processing Systems logo](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Logos/logo-NIPS.png?width=192&height=86&name=logo-NIPS.png) ## Recent **Publications** Our team contributes to the global scientific discourse through peer-reviewed journals and technical papers. Explore methodologies and data-driven insights behind our advancements. ![AI overlay on pedestrians in an urban setting with trees and structures](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Images/AI%20overlay%20on%20pedestrians%20in%20an%20urban%20setting%20with%20trees%20and%20structures.webp?width=1200&height=900&name=AI%20overlay%20on%20pedestrians%20in%20an%20urban%20setting%20with%20trees%20and%20structures.webp) [Read Now ](https://ai.sony/publications/beyond-rgb-adaptive-parallel-processing-for-raw-object-detection?hsLang=en) Beyond RGB Adaptive Parallel Processing for RAW Object Detection ![Three vertical panels with a pixelated design, left shows pink and white hues, center is grayscale with soft patterns, and right features a darker grayscale gradient](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Images/Three%20vertical%20panels%20with%20a%20pixelated%20design%2c%20left%20shows%20pink%20and%20white%20hues%2c%20center%20is%20grayscale%20with%20soft%20patterns%2c%20and%20right%20features%20a%20darker%20grayscale%20gradient.webp?width=1200&height=900&name=Three%20vertical%20panels%20with%20a%20pixelated%20design%2c%20left%20shows%20pink%20and%20white%20hues%2c%20center%20is%20grayscale%20with%20soft%20patterns%2c%20and%20right%20features%20a%20darker%20grayscale%20gradient.webp) [Read Now ](https://ai.sony/publications/noise-modeling-in-one-hour-minimizing-preparation-efforts-for-self-supervised-low-light-raw-image-denoising?hsLang=en) Noise Modeling in One Hour Minimizing Preparation Efforts for Self-supervised Low-Light RAW Image Denoising ![Futuristic layered glass blocks float above a bright surface, each containing glowing particles](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Images/Futuristic%20layered%20glass%20blocks%20float%20above%20a%20bright%20surface%2c%20each%20containing%20glowing%20particles.webp?width=1200&height=900&name=Futuristic%20layered%20glass%20blocks%20float%20above%20a%20bright%20surface%2c%20each%20containing%20glowing%20particles.webp) [Read Now ](https://ai.sony/publications/stretching-each-dollar-diffusion-training-from-scratch-on-a-micro-budget?hsLang=en) Stretching Each Dollar Diffusion Training from Scratch on a Micro-Budget ![Blurred background of a busy street with colorful stick figure overlays on moving pedestrians](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Images/Blurred%20background%20of%20a%20busy%20street%20with%20colorful%20stick%20figure%20overlays%20on%20moving%20pedestrians.webp?width=1200&height=900&name=Blurred%20background%20of%20a%20busy%20street%20with%20colorful%20stick%20figure%20overlays%20on%20moving%20pedestrians.webp) [Read Now ](https://ai.sony/publications/argus-a-compact-and-versatile-foundation-model-for-vision?hsLang=en) Argus A Compact and Versatile Vision Foundation Model ## **Latest Updates** from Sony AI Explore our latest resources to stay informed on our progress in advancing innovation around vision and sensing. ![Side profile of a human head with digital circuits and glowing pink technology integrated into it](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Images/Side%20profile%20of%20a%20human%20head%20with%20digital%20circuits%20and%20glowing%20pink%20technology%20integrated%20into%20it.webp?width=1200&height=900&name=Side%20profile%20of%20a%20human%20head%20with%20digital%20circuits%20and%20glowing%20pink%20technology%20integrated%20into%20it.webp) [Read Now ](https://www.youtube.com/watch?v=MMhstAKIIY4&t=1s) Video COALA: A Practical and Vision-Centric Federated Learning Platform [ Blog Sights on AI: Lingjuan Lyu Discusses Her Career in Privacy-Preserving AI and Staying Inspired As The AI Landscape Has Advanced ](https://ai.sony/blog/sights-on-ai-lingjuan-lyu-discusses-her-career-in-privacy-preserving-ai-and-staying-inspired-as-the-ai-landscape-has-advanced?hsLang=en) [ Blog Research That Scales, Adapts, and Creates: Spotlighting Sony AI at CVPR 2025 ](https://ai.sony/blog/research-that-scales-adapts-and-creates-spotlighting-sony-ai-at-cvpr-2025?hsLang=en) [ Blog Sights on AI: Expanding Human Perception Through AI – A Conversation with Daisuke Iso ](https://ai.sony/blog/sights-on-ai-expanding-human-perception-through-ai-a-conversation-with-daisuke-iso?hsLang=en) [ Blog Unleashing the Potential of Federated Learning with COALA ](https://ai.sony/blog/unleashing-the-potential-of-federated-learning-with-coala?hsLang=en) ![A woman with glasses and a pink jacket smiles while using a laptop on a dark sofa](https://ai.sony/hs-fs/hubfs/SonyAI_2026/Images/A%20woman%20with%20glasses%20and%20a%20pink%20jacket%20smiles%20while%20using%20a%20laptop%20on%20a%20dark%20sofa.webp?width=2000&height=1125&name=A%20woman%20with%20glasses%20and%20a%20pink%20jacket%20smiles%20while%20using%20a%20laptop%20on%20a%20dark%20sofa.webp) ## Shape the Future of AI with Sony AI [View Open Roles ](https://ai.sony/join-us?hsLang=en)