Summary of Find N’ Propagate: Open-vocabulary 3d Object Detection in Urban Environments, by Djamahl Etchegaray and Zi Huang and Tatsuya Harada and Yadan Luo
Find n’ Propagate: Open-Vocabulary 3D Object Detection in Urban Environments
by Djamahl Etchegaray, Zi Huang, Tatsuya Harada, Yadan Luo
First submitted to arxiv on: 20 Mar 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper tackles the limitations of current LiDAR-based 3D object detection systems, which are restricted by a limited vocabulary and high annotation costs. It proposes open-vocabulary learning using pre-trained vision-language models with multi-sensor data to capture novel instances in urban environments. The authors design and benchmark four potential solutions as baselines, categorizing them into top-down or bottom-up approaches based on input data strategies. These methods are effective but exhibit limitations, such as missing novel objects or applying rigorous priors that bias detections towards camera-proximal objects. To overcome these limitations, the paper introduces a universal Find n’ Propagate approach for 3D open-vocabulary tasks, maximizing novel object recall and propagating detection capability to distant areas. The authors utilize a greedy box seeker, cross alignment, density ranker, and remote simulator to alleviate bias towards camera-proximal objects. Extensive experiments demonstrate a 53% improvement in novel recall across diverse settings, vision-language models, and detectors, with up to a 3.97-fold increase in Average Precision for novel object classes. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about finding new ways to detect objects using LiDAR technology in urban environments. The current methods have limitations, such as only being able to recognize certain types of objects and being very expensive to train. Researchers are trying to use pre-trained models that combine vision and language to improve object detection. They tested different approaches, but they all had some flaws. To fix these problems, the researchers came up with a new method called Find n’ Propagate, which helps detect new objects and improves accuracy. The new approach uses special techniques to reduce bias in the results. The experiments showed that this new method is much better than the old ones, especially when it comes to detecting new types of objects. |
Keywords
» Artificial intelligence » Alignment » Object detection » Precision » Recall
