DR-NTU (Data) (Nanyang Technological University)
DR-NTU (Data) (Nanyang Technological University)Not a member yet
1955 research outputs found
Sort by
Video Diffusion Models are Training-free Motion Interpreter and Controller
No full textVideo generation primarily aims to model authentic and customized motion across frames, making understanding and controlling the motion a crucial topic. Most diffusion-based studies on video motion focus on motion customization with training-based paradigms, which, however, demands substantial training resources and necessitates retraining for diverse models. Crucially, these approaches do not explore how video diffusion models encode cross-frame motion information in their features, lacking interpretability and transparency in their effectiveness. To answer this question, this paper introduces a novel perspective to understand, localize, and manipulate motion-aware features in video diffusion models. Through analysis using Principal Component Analysis (PCA), our work discloses that robust motion-aware feature already exists in video diffusion models. We present a new MOtion FeaTure (MOFT) by eliminating content correlation information and filtering motion channels. MOFT provides a distinct set of benefits, including the ability to encode comprehensive motion information with clear interpretability, extraction without the need for training, and generalizability across diverse architectures. Leveraging MOFT, we propose a novel training-free video motion control framework. Our method demonstrates competitive performance in generating natural and faithful motion, providing architecture-agnostic insights and applicability in a variety of downstream tasks
Replication Data for: Towards Explainable In-the-Wild Video Quality Assessment: A Database and a Language-Prompted Approach
No full textA large-scale in-the-wild VQA database, named Maxwell, created to gather more than two million human opinions across 13 specific quality-related factors, including technical distortions e.g. noise, flicker and aesthetic factors e.g. contents
Dry powder microneedle-enabled transdermal anti-inflammatory therapy for obesity, diabetes, hyperlipidemia, and fatty liver
No full textObese white adipose tissue (WAT) is characterized by hypoxia, oxidative stress, and inflammation, which are the key drivers of various deleterious diseases. Herein, we demonstrate a new strategy to directly induce ameliorative remodeling of obese subcutaneous WAT (sWAT). Manganese dioxide nanoparticles, which can directly react with hydrogen peroxide and produce oxygen, and has nanocatalytic abilities mimicking superoxide dismutase and catalase, is transdermally delivered together with a natural antioxidant resveratrol, leading to reduction of oxidative stress, hypoxia, and consequently suppression of inflammation in obese sWAT. The localized treatment not only leads to remodeling of the targeted sWAT and large reduction of its mass, but also improves whole-body metabolism as evidenced by total relief of diabetes, and significant decrease of visceral fat, liver fat, hyperlipidemia, and systemic inflammation. For self-administrable and minimally-invasive transdermal delivery, a new type of microneedle is designed, which is made purely by dry powders of the therapeutics and offers a loading capacity 2 orders higher than the conventional microneedles. Moreover, the intricate signaling pathways underlying this transdermal anti-inflammatory therapy are revealed
Replication Data for: Characterization of shrink film properties for rapid microfluidics lab-on-chip fabrication
No full textReplication data for "Characterization of shrink film properties for rapid microfluid-ics lab-on-chip fabrication
FreeInit: Bridging Initialization Gap in Video Diffusion Models
No full textThough diffusion-based video generation has witnessed rapid progress, the inference results of existing models still exhibit unsatisfactory temporal consistency and unnatural dynamics. In this paper, we delve deep into the noise initialization of video diffusion models, and discover an implicit training-inference gap that attributes to the unsatisfactory inference quality. Our key findings are: 1) the spatial-temporal frequency distribution of the initial noise at inference is intrinsically different from that for training, and 2) the denoising process is significantly influenced by the low-frequency components of the initial noise. Motivated by these observations, we propose a concise yet effective inference sampling strategy, FreeInit, which significantly improves temporal consistency of videos generated by diffusion models. Through iteratively refining the spatial-temporal low-frequency components of the initial latent during inference, FreeInit is able to compensate the initialization gap between training and inference, thus effectively improving the subject appearance and temporal consistency of generation results. Extensive experiments demonstrate that FreeInit consistently enhances the generation quality of various text-to-video diffusion models without additional training or fine-tuning
Replication Data for: Automating Urban Soundscape Enhancements with AI: In-situ Assessment of Quality and Restorativeness in Traffic-Exposed Residential Areas
No full text
This dataset contains the following data tables:
amss_insitu_participant_data.csv: Participant responses to site evaluation
comb_obj_insitu_session_data.csv: Per session combined objective acoustic measurement and subjective response data
The variables contained in each data table are detailed below:
amss_insitu_participant_data.csv
pID: [numeric] Unique participant identification
site: [Factor w/ 3 levels "GND", "ROOF", "MP"] Variable describing the site, where "GND" refers to the pavilion on the ground floor, "ROOF" is the pavilion in the rooftop garden, and "MP" is the meeting point
condition: [Factor w/ 2 levels "AMB", "AMSS"] Variable describing the condition
order: [Factor w/ 2 levels "18", "81"] Variable describing the order in which the participant assesses the evaluation sites. "18" indicates that a participant evaluated the soundscape at "GND" first followed by "ROOF" and vice versa for "81".
partGrp: [Factor w/ 2 levels "multi", "single"] Variable describing the number of participants in the session for each participant, where "single" indicates that the participant conducted the experiment alone and "multi" indicates that more than one person was present during
dom_noise: [numeric] Normalised score for the dominance of noise sources (DOMNoi)
dom_human: [numeric] Normalised score for the dominance of human sound sources (DOMHum)
dom_natural: [numeric] Normalised score for the dominance of natural sound sources (DOMNat)
p: [numeric] Normalised score for the pleasant attribute in the perceived affective quality scale as per Method A questionnaire in ISO 12913-2 (rp)
e: [numeric] Normalised score for the eventful attribute in the perceived affective quality scale as per Method A questionnaire in ISO 12913-2 (re)
ch: [numeric] Normalised score for the chaotic attribute in the perceived affective quality scale as per Method A questionnaire in ISO 12913-2 (rch)
v: [numeric] Normalised score for the vibrant attribute in the perceived affective quality scale as per Method A questionnaire in ISO 12913-2 (rv)
u: [numeric] Normalised score for the uneventful attribute in the perceived affective quality scale as per Method A questionnaire in ISO 12913-2 (ru)
ca: [numeric] Normalised score for the calm attribute in the perceived affective quality scale as per Method A questionnaire in ISO 12913-2 (rca)
a: [numeric] Normalised score for the annoying attribute in the perceived affective quality scale as per Method A questionnaire in ISO 12913-2 (ra)
m: [numeric] Normalised score for the monotonous attribute in the perceived affective quality scale as per Method A questionnaire in ISO 12913-2 (rm)
overall: [numeric] Normalised score for the overall quality of the acoustic environment as per Method A questionnaire in ISO 12913-2 (OSQ)
appropriate: [numeric] Normalised score for the appropriateness of the acoustic environment as per Method A questionnaire in ISO 12913-2 (APPR)
loudness: [numeric] Normalised score for the overall loudness of the acoustic environment (PLN)
ISOPL: [numeric] Derived normalised score for the ``Pleasantness'' score calculated using equation A.1 in ISO 12913-3 (ISOPL)
ISOEV: [numeric] Derived normalised score for the ``Eventfulness'' score calculated using equation A.1 in ISO 12913-3 (ISOEV)
PRSSFas: [numeric] Derived normalised score for the ``Fascination'' dimension in the Perceived Restorativeness Soundscape Scale (PRSSFas)
PRSSBA: [numeric] Derived normalised score for the ``Being-Away'' dimension in the Perceived Restorativeness Soundscape Scale (PRSSBA)
PRSSCom: [numeric] Derived normalised score for the ``Compatibility'' dimension in the Perceived Restorativeness Soundscape Scale (PRSSCom)
PRSSEC: [numeric] Derived normalised score for the ``Extent-Coherence'' dimension in the Perceived Restorativeness Soundscape Scale (PRSSEC)
PRSSES: [numeric] Derived normalised score for the ``Extent-Scope'' dimension in the Perceived Restorativeness Soundscape Scale (PRSSES)
PosAff: [numeric] Derived normalised score for positive affect in the I-PANAS-SF scale (PA)
NegAff: [numeric] Derived normalised score for negative affect in the I-PANAS-SF scale (NA)
pss: [numeric] Derived normalised score for the 10-item perceived stress scale (PSS-10)
wnss: [numeric] Derived normalised score for the 21-item Weinstein noise sensitivity scale (INS)
wbi: [numeric] Derived normalised score for the 5-item WHO well-being index (WHO-5)
panas_*: [numeric] Individual raw ratings of the I-PANAS-SF questionnaire
annoy-*: [numeric] Individual raw ratings of the ISO 15666 questionnaire on annoyance
gender: [Factor w/ 2 levels "Male", "Female"] reported gender
occupation: [chr] reported occupation
occupation_other: [chr] self-reported occupation for "other" category
age: [numeric] reported age
pss*: [numeric] Individual raw ratings of the Perceived Stress Scale questionnaire
wnss*: [numeric] Individual raw ratings of the Weinstein Noise Sensitivity Scale questionnaire
who*: [numeric] Individual raw ratings of the WHO Well-being Index questionnaire
comb_obj_insitu_session_data.csv
date: [character] Date in YYYYMMDD format
sessionTime: [numeric] The hour of date when the evaluation session started
site: [Factor w/ 2 levels "GND", "ROOF"] Variable describing the site
condition: [Factor w/ 2 levels "AMB", "AMSS"] Variable describing the condition
L[Aeq]: [numeric] The C-weighted equivalent sound pressure level for 10-min listening period of the session at date and sessionTime
L[Ceq]: [numeric] The C-weighted equivalent sound pressure level for 10-min listening period of the session at date and sessionTime
N[95]: [numeric] The 95 % exceedance levels of psychoacoustic loudness for 10-min listening period of the session at date and sessionTime
pair: [numeric] A combination of condition--site for plotting purposes
attribute: [numeric] The collated subjective attributes of ISOPL, OSQ, PRSSFas, PRSSBA, and PRSSCom
score: [numeric] The hour of date when the evaluation session started
enviro_session_data.csv
startDate: [POSIXlt] The local POSIXlt datetime indicating the start of the 10-min listening period of the session
endDate: [POSIXlt] The local POSIXlt datetime indicating the end of the 10-min listening period of the session
condition: [Factor w/ 2 levels "AMB", "AMSS"] Variable describing the condition
temperature: [numeric] Variable describing the temperature in °C
humidity: [numeric] Variable describing the humidity in %RH
lux: [numeric] Variable describing the luminance in lux
windspeed: [numeric] Variable describing the wind speed in km/h
24h_psi: [numeric] Variable describing the 24 hour Pollution Standards Index
pm25: [numeric] Variable describing the 1-hour PM2.5 reading in µg/m3
predict_session_data.csv
timestamp: [POSIXlt] The local POSIXlt datetime indicating the time at which the 30-s long masker was played at the ROOF pavilion
predictions: [numeric] Variable describing the masker that was chosen based on the highest predicted ISOPL increase
condition: [Factor w/ 1 level "AMSS"] Variable describing the condition
fullData.RData
amss_insitu_participant_data: [data.frame] R S3 tibble data.frame containing the data in amss_insitu_participant_data.csv
comb_obj_insitu_session_data: [data.frame] R S3 tibble data.frame containing the data in comb_obj_insitu_session_data.csv
enviro_session_data: [data.frame] R S3 tibble data.frame containing the data in enviro_session_data.csv
predict_session_data: [data.frame] R S3 tibble data.frame containing the data in predict_session_data.csv
</ul
An investigation of the structural and electronic origins of enhanced chemical looping air separation performance of B-site substituted SrFe1-xCoxO3-δ perovskites
No full textThis dataset provides the research data for the manuscript "An investigation of the structural and electronic origins of enhanced chemical looping air separation performance of B-site substituted SrFe1-xCoxO3-δ perovskites"
Related Data for: All-Round Ionic Liquids for Shuttle-Free Zinc-Iodine Battery
No full textThe practical implementation of aqueous zinc-iodine batteries (ZIBs) is hindered by the rampant Zn dendrites growth, parasite corrosion, and polyiodide shuttling. In this work, ionic liquid EMIM[OAc] is employed as an all-round solution to mitigate challenges on both the Zn anode and the iodine cathode side. First, the EMIM+ embedded lean-water inner Helmholtz plane (IHP) and inert solvation sheath modulated by OAc− effectively repels H2O molecules away from the Zn anode surface. The preferential adsorption of EMIM+ on Zn metal facilitates uniform Zn nucleation via a steric hindrance effect. Second, EMIM+ can reduce the polyiodide shuttling by hindering the iodine dissolution and forming an EMIM+-I3− dominated phase. These effects holistically enhance the cycle life, which is manifested by both Zn || Zn symmetric cells and Zn-I2 full cells. ZIBs with EAc deliver a capacity decay rate of merely 0.01 ‰ per cycle after over 18,000 cycles at 4 A g−1, and lower self-discharge and better calendar life than the ZIBs without ionic liquid EAc additive
Data for: Haptic Manipulation of 3D Scans for Geometric Feature Enhancement (Refining 3D scan of a surface by manual haptic exploration)
No full textObject localisation, shape estimation, feature detection are important sub-tasks in overarching applications like industrial tooling, assembly, manufacturing and re-manufacturing which might require robotic automation through manipulation. 3D vision techniques provide robust solutions but fall short in capturing all types of 3D features. This work details a novel haptic framework that allows to refine a 3D scan of a surface using actual forces recorded by virtue of haptic feedback due to surface tool interaction. The current work is limited to interaction with rigid objects using pen-like end-effectors. The proposed framework is presented qualitatively followed by the experimental implementation for human-surface interaction. A 3D scan of a surface is obtained as a polygonal mesh and haptic exploration is performed to adapt the mesh to capture the geometry of absent features like holes and edges. Finally, conclusions from this work are drawn and future work is discussed
Basal Cell Carcinoma by Reflectance Confocal Microscopy images
No full textClassification on reflectance confocal microscopy images of basal cell carcinoma