-
Towards Multi-domain Face Landmark Detection with Synthetic Data from Diffusion model
Authors:
Yuanming Li,
Gwantae Kim,
Jeong-gi Kwak,
Bon-hwa Ku,
Hanseok Ko
Abstract:
Recently, deep learning-based facial landmark detection for in-the-wild faces has achieved significant improvement. However, there are still challenges in face landmark detection in other domains (e.g. cartoon, caricature, etc). This is due to the scarcity of extensively annotated training data. To tackle this concern, we design a two-stage training approach that effectively leverages limited data…
▽ More
Recently, deep learning-based facial landmark detection for in-the-wild faces has achieved significant improvement. However, there are still challenges in face landmark detection in other domains (e.g. cartoon, caricature, etc). This is due to the scarcity of extensively annotated training data. To tackle this concern, we design a two-stage training approach that effectively leverages limited datasets and the pre-trained diffusion model to obtain aligned pairs of landmarks and face in multiple domains. In the first stage, we train a landmark-conditioned face generation model on a large dataset of real faces. In the second stage, we fine-tune the above model on a small dataset of image-landmark pairs with text prompts for controlling the domain. Our new designs enable our method to generate high-quality synthetic paired datasets from multiple domains while preserving the alignment between landmarks and facial features. Finally, we fine-tuned a pre-trained face landmark detection model on the synthetic dataset to achieve multi-domain face landmark detection. Our qualitative and quantitative results demonstrate that our method outperforms existing methods on multi-domain face landmark detection.
△ Less
Submitted 23 January, 2024;
originally announced January 2024.
-
Dis/Immersion in Mindfulness Meditation with a Wandering Voice Assistant
Authors:
Bonhee Ku,
Tatsuya Itagaki,
Katie Seaborn
Abstract:
Mindfulness meditation is a validated means of hel** people manage stress. Voice-based virtual assistants (VAs) in smart speakers, smartphones, and smart environments can assist people in carrying out mindfulness meditation through guided experiences. However, the common fixed location embodiment of VAs makes it difficult to provide intuitive support. In this work, we explored the novel embodime…
▽ More
Mindfulness meditation is a validated means of hel** people manage stress. Voice-based virtual assistants (VAs) in smart speakers, smartphones, and smart environments can assist people in carrying out mindfulness meditation through guided experiences. However, the common fixed location embodiment of VAs makes it difficult to provide intuitive support. In this work, we explored the novel embodiment of a "wandering voice" that is co-located with the user and "moves" with the task. We developed a multi-speaker VA embedded in a yoga mat that changes location along the body according to the meditation experience. We conducted a qualitative user study in two sessions, comparing a typical fixed smart speaker to the wandering VA embodiment. Thick descriptions from interviews with twelve people revealed sometimes simultaneous experiences of immersion and dis-immersion. We offer design implications for "wandering voices" and a new paradigm for VA embodiment that may extend to guidance tasks in other contexts.
△ Less
Submitted 21 April, 2023;
originally announced April 2023.
-
RankBooster: Visual Analysis of Ranking Predictions
Authors:
Abishek Puri,
Bon Kyung Ku,
Yong Wang,
Huamin Qu
Abstract:
Ranking is a natural and ubiquitous way to facilitate decision-making in various applications. However, different rankings are often used for the same set of entities, with each ranking method placing emphasis on different factors. These factors can also be multi-dimensional in nature, compounding the problem. This complexity can make it challenging for an entity which is being ranked to understan…
▽ More
Ranking is a natural and ubiquitous way to facilitate decision-making in various applications. However, different rankings are often used for the same set of entities, with each ranking method placing emphasis on different factors. These factors can also be multi-dimensional in nature, compounding the problem. This complexity can make it challenging for an entity which is being ranked to understand what they can do to improve their rankings, and to analyze the effect of changes in various factors to their overall rank. In this paper, we present RankBooster, a novel visual analytics system to help users conveniently investigate ranking predictions. We take university rankings as an example and focus on hel** universities to better explore their rankings, where they can compare themselves to their rivals in key areas as well as overall. Novel visualizations are proposed to enable efficient analysis of rankings, including a Scenario Analysis View to show a high-level summary of different ranking scenarios, a Relationship View to visualize the influence of each attribute on different indicators and a Rival View to compare the ranking of a university and those of its rivals. A case study demonstrates the usefulness and effectiveness of RankBooster in facilitating the visual analysis of ranking predictions and hel** users better understand their current situation.
△ Less
Submitted 14 April, 2020;
originally announced April 2020.
-
Pulse: Toward a Smart Campus by Communicating Real-time Wi-Fi Access Data
Authors:
Aoyu Wu,
Bon Kyung Ku,
Furui Cheng,
Xinhuan Shu,
Abishek Puri,
Yifang Wang,
Huamin Qu
Abstract:
To enhance the mobility and convenience of the campus community, we designed and implemented the Pulse system, a visual interface for communicating the crowd information to the lay public including campus members and visitors. This is a challenging task which requires analyzing and reconciling the demands and interests for data as well as visual design among diverse target audiences. Through an it…
▽ More
To enhance the mobility and convenience of the campus community, we designed and implemented the Pulse system, a visual interface for communicating the crowd information to the lay public including campus members and visitors. This is a challenging task which requires analyzing and reconciling the demands and interests for data as well as visual design among diverse target audiences. Through an iterative design progress, we study and address the diverse preferences of the lay audiences, whereby design rationales are distilled. The final prototype combines a set of techniques such as chart junk and redundancy encoding. Initial feedback from a wide audience confirms the benefits and attractiveness of the system.
△ Less
Submitted 29 September, 2018;
originally announced October 2018.
