Sitemap
A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.
Pages
Posts
What We Have Built At Deep Render (So Far)
Published:
Deep Render was built from the ground up to develop a production ready codec. We focused on developing an AI codec with low computational complexity and high compression efficiency. We worked with a relentless product focus, efficient tooling and process driven research. After 4 years in the trenches, we have developed a codec that can achieve real-time encode and decode with over 45% BD rate saving w.r.t. SVT-AV1. In other words, the world’s first AI codec.
Future of AI based compression
Published:
Deep Render recently introduced the first AI codec into FFmpeg and VLC. This marks a significant step for AI codecs, making them readily available in tools widely used in the compression industry. Deep Render will continue to push the frontier of AI codecs by further improving compression performance, providing support for more hardware platforms and improving feature diversity.
Solving AI based compression
Published:
Deep Render recently shipped the world’s first AI codec to its customers. This result follows two years of careful research, gruelling engineering, and relentless product focus. I’ll take some time to share some thoughts on the research and engineering that went into solving AI-based compression.
Transformer and ViT dataflow notation
Published:
When reviewing the transformer and ViT literature, to get an intuitive understanding of the various model layers and how the input tokens are manipulated, I found it helpful to map out the data flow through the model in matrix notation. I couldn’t find this anywhere so I thought I’d share it.
Diffusion Decoder based Compression
Published:
I’m always super interested in expanding the fronteir of learned compression. The current methods we use are derived from VAEs (equivalence shown here) and use GANs to improve preceptual quality but in this blog I explore how diffusion models could be used for compression.
Proof: MSE makes GANs locally stable
Published:
Generative Adversarial Networks are notoriously unstable due to issues such as mode collapse and training divergence. However, in AI compression, the adversarial training is generally stable and reliable without the neccessary tricks such as gradient penalty and adding noise to our input samples. I found this intriguing so I set out to explore why.
Cross-Entropy, KL and NLL are the same objective in AI compression
Published:
In end-to-end learned compression, we need to model the distribution of data, so we can entropy encode it. The training loss we use is the cross-entropy between the unknown true data distribution and our model distribution which in a simplest case is a fully factorized distribution of standard normals. There is often some confusion about the objective used in compression so I thought I’d use this post to clarify it. I’ll show that for data sampled from the true distribution \(p\) and a parametric model \(q_\theta\) we learn, the cross-entropy, the Kullback–Leibler divergence and the negative log-likelihood are optimization-equivalent objectives. In practice this means we are doing maximum likelihood estimation (MLE) and simultaneously minimizing expected code length.
portfolio
Portfolio item number 1
Published:
Short description of portfolio item number 1
Portfolio item number 2
Published:
Short description of portfolio item number 2
publications
Paper Title Number 1
Published in Journal 1, 2009
This paper is about the number 1. The number 2 is left for future work.
Recommended citation: Your Name, You. (2009). "Paper Title Number 1." Journal 1. 1(1).
Download Paper | Download Slides
Paper Title Number 2
Published in Journal 1, 2010
This paper is about the number 2. The number 3 is left for future work.
Recommended citation: Your Name, You. (2010). "Paper Title Number 2." Journal 1. 1(2).
Download Paper | Download Slides
Paper Title Number 3
Published in Journal 1, 2015
This paper is about the number 3. The number 4 is left for future work.
Recommended citation: Your Name, You. (2015). "Paper Title Number 3." Journal 1. 1(3).
Download Paper | Download Slides
Paper Title Number 4
Published in GitHub Journal of Bugs, 2024
This paper is about fixing template issue #693.
Recommended citation: Your Name, You. (2024). "Paper Title Number 3." GitHub Journal of Bugs. 1(3).
Download Paper
talks
AI Image Compression Demo at Imperial College London
Published:
In this video, I demonstate the first ever end-to-end AI Image compression system we build at Deep Render. This system was able to outperform BPG. This talk and demo was given at Imperial College London to a group of PhD students, Professors and CS students.
Overview of cuDNN algorithms
Published:
In this video, I provide an overview of various cuDNN algorithms by reviewing the 2019 paper by Marc Jorda et al.
Implicit Density Estimation
Published:
In this talk, I provide an overview of the main methods used in implicit density estimation. These methods prove to be very useful for many domains, including AI-based video compression.
Normalizing flows and VAEs
Published:
In this video, Chris Finlay and I provide an overview of VAEs and normalizing flows through presenting the SURVAE paper which aims bridge the gap between them.
Intel Innovation Demo
Published:
In this video, we were invited by Intel CEO Pat Gelsinger to present our AI based video compression at the Intel Innovation conference in Santa Clara.
A Conversation on AI Codecs with Jona Ballé
Published:
AI codecs are replacing traditional block-based compression algorithms developed by standards bodies such as MPEG and ITU with fully AI-based encoder-decoder models. These AI models are trained end-to-end on video data and already outperform traditional codecs. Undoubtedly, all future codecs will be AI-powered.
Intel Innovation Award
Published:
In this video, we presented our AI based video compression for the Intel Innovatation conference in San Jose.
Primer on AI based compression
Published:
In this video, I provide a primer on end-to-end AI based compression. I outline the training, inference and deployment process for AI codecs. This talk is aimed at video codec experts looking to enter the AI codec space.
World’s first AI codec
Published:
In this video, we introduce the world’s first AI codec into FFMPEG and VLC. We highlight the benifits of AI codecs and show their readiness through a demonstration on current hardware.
teaching
Teaching experience 1
Undergraduate course, University 1, Department, 2014
This is a description of a teaching experience. You can use markdown like any other post.
Teaching experience 2
Workshop, University 1, Department, 2015
This is a description of a teaching experience. You can use markdown like any other post.