Technology

and Research

Take advantage of the power of AI

    Unlock the full potential of artificial intelligence with our cutting-edge technology.

    Our innovative solution leverages the power of Deep Neural Networks to perform advanced video enhancements, delivering superior results that exceed traditional methods.

    Our team of experienced Deep Neural Network engineers has developed Machine Learning models based on Generative Adversarial Networks (GANs) to achieve state-of-the-art image enhancement.

    With our technology, you can enjoy high-quality video resolution and image enhancement that is unmatched by other methods.

Hyper-optimized mobile enhancement

    Small Pixels has developed a cutting-edge technology that leverages hyper-optimized mobile enhancement techniques.

    The technology is available in a lightweight format and allows for real-time video enhancement on a variety of platforms including mobile devices, PCs, and SmartTVs, even those with limited computing capabilities.

    The innovation promises to deliver high-quality video resolution and image enhancement without straining the resources of low-powered devices, thereby expanding the market for advanced video features.

Scientific Validation

VMAF improvement

Video Multimethod Assessment Fusion (VMAF) is an objective video quality metric, jointly developed by the visionary minds at Netflix and the prestigious University of Southern California (USC). This technology utilizes a reference and distorted video sequence to accurately predict the subjective video quality.

With Small Pixels AI, the VMAF performance scales unprecedented heights, rendering up to a remarkable 44% improvement over the original video, while maintaining the same bit-rate.

Bit-rate optimization

Not only does this cutting-edge technology enhance the visual quality of videos, but it also can astoundingly act in reducing the bitrate requirements while maintaining a consistent viewing experience. The application of this advanced technique has resulted in up to 50% reduction in bit-rate and 33% average reduction, without any compromise in terms of the video's perceptible quality.

Distinctive Features

  • Pipeline agnostic

    This exceptional technology is designed to be pipeline agnostic, allowing it to be seamlessly integrated into any existing pipeline using a simple software add-on. Moreover, it leverages hardware acceleration to deliver optimal performance when an AI engine is available, enabling unparalleled efficiency.

  • Codec agnostic

    With its state-of-the-art algorithms, this remarkable technology is capable of delivering top-tier results with currently available codecs. Furthermore, it's future-proofed to accommodate next-gen codecs, ensuring that it remains relevant and effective for years to come.

  • Content agnostic

    This extraordinary technology is content agnostic and has the capability to remove compression artifacts and perform frame rescaling on any type of video content, regardless of its format, resolution, or encoding. Its versatility and adaptability make it the ideal solution for a wide range of video processing needs.

  • Context adaptive

    This cutting-edge technology specializes in adapting to specific content or distortion effects, enabling it to tackle even the most challenging and complex situations with ease. Its context-adaptive feature empowers it to deliver exceptional performance by precisely addressing the distortions and preserving the quality of the original video.

Research

Small Pixels is a spin-off of Media Integration and Communication Center (MICC) of University of Firenze, Italy. The MICC works as an interdisciplinary center for advanced research in the fields of artificial intelligence, computer vision, multimedia technologies applied to smart environments, natural interaction, Internet Based Applications and collective intelligence. Our products directly stem from top quality research activity and strong connections with academic research at the national and international level.

Scientific Publications

M. Bertini, L. Galteri, L. Seidenari, T. Uricchio, A. Del Bimbo, Proceedings of the 1st Mile-High Video Conference2022.
L. Galteri, L. Seidenari, T. Uricchio, M. Bertini, A. Del Bimbo, Multi-faceted Deep Learning2021.
F. Mameli, M. Bertini, L. Galteri, A. Del Bimbo, International Conference on Pattern Recognition2021.
F. Vaccaro, M. Bertini, T. Uricchio, A. Del Bimbo, Proceedings of the 1st Mile-High Video Conference2022.
L. Galteri, M. Bertini, L. Seidenari, T. Uricchio, A. Del Bimbo, Proceedings of the 28th ACM International Conference on Multimedia2020.
L. Galteri, L. Seidenari, M. Bertini, A. Del Bimbo, International Conference on Computer Analysis of Images and Patterns2019.
F. Mameli, M. Bertini, L. Galteri, A. Del Bimbo, Proceedings of the 28th ACM International Conference on Multimedia2020.
L. Galteri, L. Seidenari, T. Uricchio, M. Bertini, A. Del Bimbo, IOP Conference Series: Materials Science and Engineering2020.
L. Galteri, C. Ferrari, G. Lisanti, S. Berretti, A. Del Bimbo, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition2019.
L. Galteri, L. Seidenari, M. Bertini, A. Del Bimbo, IEEE Transactions on Multimedia vol. 212019.
L. Galteri, C. Ferrari, G. Lisanti, S. Berretti, A. Del Bimbo, Computer Vision and Image Understanding vol. 1852019.
L. Galteri, L. Seidenari, M. Bertini, T. Uricchio, A. Del Bimbo, Proceedings of the 27th ACM International Conference on Multimedia2019.
L. Galteri, C. Ferrari, G. Lisanti, S. Berretti, A. Del Bimbo, 14th IEEE International Conference on Automatic Face & Gesture Recognition2019.
L. Galteri, L. Seidenari, M. Bertini, T. Uricchio, A. Del Bimbo, Proceedings of the IEEE International Conference on Computer Vision2017.