Recent advances in visual generation have made significant strides in producing content of exceptional quality. However, most methods suffer from a fundamental problem - a bottleneck of inference computational efficiency. Most of these algorithms involve multiple passes over a transformer model to generate tokens or denoise inputs. However, the model size is kept consistent throughout all iterations, which makes it computationally expensive. In this work, we aim to address this issue primarily through two key ideas - (a) not all parts of the generation process need equal compute, and we design a decode time model scaling schedule to utilize compute effectively, and (b) we can cache and reuse some of the intermediate computation. Combining these two ideas leads to using smaller models to process more tokens while large models process fewer tokens. These different-sized models do not increase the parameter size, as they share parameters. We rigorously experiment with ImageNet256$\times$256 , UCF101, and Kinetics600 to showcase the efficacy of the proposed method for image/video generation and frame prediction. Our experiments show that with almost $3\times$ less compute than baseline, our model obtains competitive performance.

Graphic designs are an effective medium for visual communication. They range from greeting cards to corporate flyers and beyond. Off-late, machine learning techniques are able to generate such designs, which accelerates the rate of content production. An automated way of evaluating their quality becomes critical. Towards this end, we introduce Design-o-meter, a data-driven methodology to quantify the goodness of graphic designs. Further, our approach can suggest modifications to these designs to improve its visual appeal. To the best of our knowledge, Design-o-meter is the first approach that scores and refines designs in a unified framework despite the inherent subjectivity and ambiguity of the setting. Our exhaustive quantitative and qualitative analysis of our approach against baselines adapted for the task (including recent Multimodal LLM-based approaches) brings out the efficacy of our methodology. We hope our work will usher more interest in this important and pragmatic problem setting. Project Page: sahilg06.github.io/Design-o-meter.

Lip synchronization and talking face generation have gained a specific interest from the research community with the advent and need of digital communication in different fields. Prior works propose several elegant solutions to this problem. However, they often fail to create realistic-looking videos that account for people's expressions and emotions. To mitigate this, we build a talking face generation framework conditioned on a categorical emotion to generate videos with appropriate expressions, making them more real-looking and convincing. With a broad range of six emotions i.e., anger, disgust, fear, happiness, neutral, and sad, we show that our model generalizes across identities, emotions, and languages.

Binocular disparity allows interesting visual effects visible only to people with stereoscopic 3D displays. Here, we studied and applied one such effect, binocular luster, to the application of digital colorblind aids with active shutter 3D. We developed two prototype techniques, ColorBless and PatternBless, to investigate the effectiveness of such aids and to explore the potential applications of a luster effect in stereoscopic 3D beyond highlighting. User studies and interviews revealed that luster-based aids were fast and required lower cognitive effort than existing aids and were preferred over other aids by the majority of colorblind participants. We infer design implications of a luster effect from the study and propose potential applications in augmented visualization.

Journal Article Shared Input Multimodal Mobile Interfaces: Interaction Modality Effects on Menu Selection in Single-Task and Dual-Task Environments Get access Shengdong Zhao, Shengdong Zhao * 1Department of Computer Science, National University of Singapore, 13 Computing Drive, Computing 2, #01-04, Singapore 117417 *Corresponding author: [email protected] Search for other works by this author on: Oxford Academic Google Scholar Duncan P. Brumby, Duncan P. Brumby 2UCL Interaction Centre, University College London, Gower Street, London WC1E 6BT, UK Search for other works by this author on: Oxford Academic Google Scholar Mark Chignell, Mark Chignell 3Knowledge Media Design Institute (KMDI), University of Toronto, 27 King's College Circle, Toronto, Ont., Canada M5S 1A1 Search for other works by this author on: Oxford Academic Google Scholar Dario Salvucci, Dario Salvucci 4Drexel University, 3141 Chestnut Street, Philadelphia, PA 19104, USA Search for other works by this author on: Oxford Academic Google Scholar Sahil Goyal Sahil Goyal 5National University of Singapore, 13 Computing Drive, Computing 2, #01-04, Singapore 117417 Search for other works by this author on: Oxford Academic Google Scholar Interacting with Computers, Volume 25, Issue 5, September 2013, Pages 386–403, https://doi.org/10.1093/iwc/iws021 Published: 06 February 2013 Article history Received: 29 December 2011 Revision received: 11 October 2012 Accepted: 13 November 2012 Published: 06 February 2013

This paper presents a verification framework that attempts to bridge the disconnect between high-level properties capturing the architectural power management strategy and the implementation of the power management control logic using low-level per-domain control signals. The novelty of the proposed framework is in demonstrating that the architectural power intent properties developed using high-level artifacts can be automatically translated into properties over low-level control sequences gleaned from UPF specifications of power domains, and that the resulting properties can be used to formally verify the global on-chip power management logic. The proposed translation uses a considerable amount of domain knowledge and is also not purely syntactic, because it requires formal extraction of timing information for the low-level control sequences. We present a tool, called POWER-TRUCTOR which enables the proposed framework, and several test cases of significant complexity to demonstrate the feasibility of the proposed framework.