Datasets

Dense annotations for texture and subtexture across synthetic images, with over 800,000 synthetic images annotated at texture and subtexture levels.

A non-photorealistic dataset made up of 1,336,272 painterly depictions of many objects in several colors and hand-drawn artistic styles, including automatically computed descriptions of their appearance.

A photorealistic dataset designed for disentanglement tasks, created by systematically combining 30 geometries, 365 measured materials, and 9 lighting conditions, totaling 98,550 synthetic images.

A dataset of physiological and behavioral signals captured in a VR visual-search experience under different levels of multisensory cognitive load and movement, including ECG, EDA, PPG, respiration, eye tracking, performance, and subjective questionnaires from 36 participants.

A dataset of eyetracking data captured in a VR audiovisual environment. Participants performed a visual search task while sometimes receiving simultaneous auditory cues. The dataset includes audiovisual cue information and participants' detection and recognition performance.

Eye and head data recorded for 37 participants, plus 14 pilot participants, performing free exploration, memory, and visual search tasks in three different 3D scenes.

A dataset of time-perception experiments recorded both in a VR audiovisual environment and with traditional displays. It includes stimuli, participant responses for three experiments, and a fast-forward recording of an experiment.

NLOS imaging datasets corresponding to the experiments in the paper, using fourth- and fifth-bounce photons to reconstruct objects around two corners and address the missing-cone problem.

D-SAV360 is composed of 50 stereoscopic and 35 monoscopic videos with ambisonic sounds. It provides gaze data from 87 participants, saliency maps, optical flow, estimated depth, and computed audio energy maps, together with a Unity capture and visualization system.

mitransient is a library that extends the Mitsuba 3 path tracer with support for transient simulations, including polarization tracking, differentiable transient rendering, frequency-space rendering, and non-line-of-sight (NLOS) data capture simulations.

y-tal (or tal) is a Python library providing utilities for non-line-of-sight imaging research, including tools for data generation, analysis, and implementations of multiple reconstruction algorithms.