This work presents OVIR-3D, a straightforward yet effective method for open-vocabulary 3D object instance retrieval without using any 3D data for training. Given a language query, the proposed method is able to return a ranked set of 3D object instance segments based on the feature similarity of the instance and the text query. This is achieved by a multi-view fusion of text-aligned 2D region proposals into 3D space, where the 2D region proposal network could leverage 2D datasets, which are more accessible and typically larger than 3D datasets. The proposed fusion process is efficient as it can be performed in real-time for most indoor 3D scenes and does not require additional training in 3D space. Experiments on public datasets and a real robot show the effectiveness of the method and its potential for applications in robot navigation and manipulation.
more »
« less
CVD Synthesis of 3D-Shaped 3D Graphene Using a 3D-Printed Nickel–PLGA Catalyst Precursor
- Award ID(s):
- 1727472
- NSF-PAR ID:
- 10356528
- Date Published:
- Journal Name:
- ACS Omega
- Volume:
- 6
- Issue:
- 43
- ISSN:
- 2470-1343
- Page Range / eLocation ID:
- 29009 to 29021
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
During the 2017 Engineering-a-Future activities at Tennessee Tech University, a set of 3D Pen activities has been organized for the female Middle School Students (5th and 6th grades) so that they learn more about the engineering fields and their practices in daily life. Various Additive Manufacturing technologies have been presented to students, and students have toured the lab facilities. During the 45-minute timeframe, students were grouped and constructed 3D art objects (butterfly, necklace, bird, and glasses) using the available 3D Pens. It was observed that the students had high enthusiasm about the new technology. This study will present the utilization of 3D Printing Pens in various practices, and report the current advantageous and disadvantageous of 3D Pen exercises detected through recent studies.more » « less
-
Challenge: Digital Imaging of root traits 3D (DIRT/3D) [1] is a software to measure 3D root traits on excavated roots crowns from the field. However, quantifying 3D root traits remains a challenge due to the unknown tradeoff between 3D root-model quality and 3D root-trait accuracy [2]. Questions: Can the 3D root model reconstruction be improved while reducing the image-capturing effort? Does improved 3D root model quality increase the accuracy of trait measurements? Evaluation: Compare reconstruction performance of five open-source 3D model reconstruction pipelines on 12 architecturally contrasting genotypes [1] of field-grown maize roots. Evaluate the accuracy of 3D root traits between the original implementation of DIRT/3D based on VisualSFM with an implementation based on COLMAP. Conclusion: The updated DIRT/3D (COLMAP) pipeline enables quicker image collection by reducing the number of images needed and reducing the human factor during image collection. The results demonstrate that the accuracy of 3D root-trait measurements remained uncompromised.more » « less