An official website of the United States government
Abstract Well-studied techniques that enhance diversity in early design concept generation require effective metrics for evaluating human-perceived similarity between ideas. Recent work suggests collecting triplet comparisons between designs directly from human raters and using those triplets to form an embedding where similarity is expressed as a Euclidean distance. While effective at modeling human-perceived similarity judgments, these methods are expensive and require a large number of triplets to be hand-labeled. However, what if there were a way to use AI to replicate the human similarity judgments captured in triplet embedding methods? In this paper, we explore the potential for pretrained Large Language Models (LLMs) to be used in this context. Using a dataset of crowdsourced text descriptions written about engineering design sketches, we generate LLM embeddings and compare them to an embedding created from human-provided triplets of those same sketches. From these embeddings, we can use Euclidean distances to describe areas where human perception and LLM perception disagree regarding design similarity. We then implement this same procedure but with descriptions written from a template that attempts to isolate a particular modality of a design (i.e., functions, behaviors, structures). By comparing the templated description embeddings to both the triplet-generated and pre-template LLM embeddings, we identify ways of describing designs such that LLM and human similarity perception better agree. We use these results to better understand how humans and LLMs interpret similarity in engineering designs.
more »
« less
A Non-Laboratory Gait Dataset of Full Body Kinematics and Egocentric Vision
Abstract In this manuscript, we describe a unique dataset of human locomotion captured in a variety of out-of-the-laboratory environments captured using Inertial Measurement Unit (IMU) based wearable motion capture. The data contain full-body kinematics for walking, with and without stops, stair ambulation, obstacle course navigation, dynamic movements intended to test agility, and negotiating common obstacles in public spaces such as chairs. The dataset contains 24.2 total hours of movement data from a college student population with an approximately equal split of males to females. In addition, for one of the activities, we captured the egocentric field of view and gaze of the subjects using an eye tracker. Finally, we provide some examples of applications using the dataset and discuss how it might open possibilities for new studies in human gait analysis.
more »
« less
- Award ID(s):
- 2024446
- PAR ID:
- 10391087
- Publisher / Repository:
- Nature Publishing Group
- Date Published:
- Journal Name:
- Scientific Data
- Volume:
- 10
- Issue:
- 1
- ISSN:
- 2052-4463
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
null (Ed.)In a human-centered intelligent manufacturing system, every element is to assist the operator in achieving the optimal operational performance. The primary task of developing such a human-centered system is to accurately understand human behavior. In this paper, we propose a fog computing framework for assembly operation recognition, which brings computing power to the data source, to achieve real-time recognition. The operator’s activity is captured using visual cameras. Instead of directly training a deep learning model from scratch, transfer learning is applied to transfer the learning abilities to our application. A worker assembly operation dataset is established, which at present contains 10 sequential operations in an assembly task of installing a desktop CNC machine. The developed model is evaluated on this dataset and achieves a recognition accuracy of 95% in the testing experiments.more » « less
-
There is a lack of datasets for visual-inertial odometry applications in Extended Reality (XR). To the best of our knowledge, there is no dataset available that is captured from an XR headset with a human as a carrier. To bridge this gap, we present a novel pose estimation dataset --- called HoloSet --- collected using Microsoft Hololens 2, which is a state-of-the-art head mounted device for XR. Potential applications for HoloSet include visual-inertial odometry, simultaneous localization and mapping (SLAM), and additional applications in XR that leverage visual-inertial data. HoloSet captures both macro and micro movements. For macro movements, the dataset consists of more than 66,000 samples of visual, inertial, and depth camera data in a variety of environments (indoor, outdoor) and scene setups (trails, suburbs, downtown) under multiple user action scenarios (walk, jog). For micro movements, the dataset consists of more than 12,000 samples of additional articulated hand depth camera images while a user plays games that exercise fine motor skills and hand-eye coordination. We present basic visualizations and high-level statistics of the data and outline the potential research use cases for HoloSet.more » « less
-
Nowadays, to assess and document construction and building performance, large amount of visual data are captured and stored through camera equipped platforms such as wearable cameras, unmanned aerial/ground vehicles, and smart phones. However, due to the nonstop fashion in recording such visual data, not all of the frames in captured consecutive footages are intentionally taken, and thus not every frame is worthy of being processed for construction and building performance analysis. Since many frames will simply have non-construction related contents, before processing the visual data, the content of each recorded frame should be manually investigated depending on the association with the goal of the visual assessment. To address such challenges, this paper aims to automatically filter construction big visual data that requires no human annotations. To overcome challenges in pure discriminative approach using manually labeled images, we construct a generative model with unlabeled visual dataset, and use it to find construction-related frames in big visual dataset from jobsites. First, through composition-based snap point detection together with domain adaptation, we filter and remove most of accidently recorded frames in the footage. Then, we create discriminative classifier trained with visual data from jobsites to eliminate non-construction related images. To evaluate the reliability of the proposed method, we have obtained the ground truth based on human judgment for each photo in our testing dataset. Despite learning without any explicit labels, the proposed method shows a reasonable practical range of accuracy, which generally outperforms prior snap point detection. Through the case studies, the fidelity of the algorithm is discussed in detail. By being able to focus on selective visual data, practitioners will spend less time on browsing large amounts of visual data; rather spend more time on looking at how to leverage the visual data to facilitate decision-makings in built environments.more » « less
-
Fortson, Lucy; Crowston, Kevin; Kloetzer, Laure; Ponti, Marisa (Ed.)Using public support to extract information from vast datasets has become a popular method for accurately labeling wildlife data in camera trap (CT) images. However, the increasing demand for volunteer effort lengthens the time interval between data collection and our ability to draw ecological inferences or perform data-driven conservation actions. Artificial intelligence (AI) approaches are currently highly effective for species detection (i.e., whether an image contains animals or not) and labeling common species; however, it performs poorly on species rarely captured in images and those that are highly visually similar to one another. To capitalize on the best of human and AI classifying methods, we developed an integrated CT data pipeline in which AI provides an initial pass on labeling images, but is supervised and validated by humans (i.e., a “human-in-the-loop” approach). To assess classification accuracy gains, we compare the precision of species labels produced by AI and HITL protocols to a “gold standard” (GS) dataset annotated by wildlife experts. The accuracy of the AI method was species-dependent and positively correlated with the number of training images. The combined efforts of HITL led to error rates of less than 10% for 73% of the dataset and lowered the error rates for an additional 23%. For two visually similar species, human input resulted in higher error rates than AI. While integrating humans in the loop increases classification times relative to AI alone, the gains in accuracy suggest that this method is highly valuable for high-volume CT surveys.more » « less
