Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
UNION: Unsupervised 3D Object Detection using Object Appearance-based Pseudo-Classes
Authors: Ted Lentsch, Holger Caesar, Dariu Gavrila
NeurIPS 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments on the nu Scenes dataset and increase the state-of-the-art performance for unsupervised 3D object discovery, i.e. UNION more than doubles the average precision to 38.4. |
| Researcher Affiliation | Academia | Ted Lentsch Holger Caesar Dariu M. Gavrila Department of Cognitive Robotics Delft University of Technology |
| Pseudocode | No | The paper describes the method pipeline in text and a figure (Figure 1), but does not include a formal pseudocode or algorithm block. |
| Open Source Code | Yes | The code is available at github.com/Ted Lentsch/UNION. |
| Open Datasets | Yes | Dataset. We evaluate our method on the challenging nu Scenes [3] dataset. This is a large-scale autonomous driving dataset for 3D perception captured in diverse weather and lighting conditions across Boston and Singapore. |
| Dataset Splits | Yes | It consists of 700, 150, and 150 scenes for training, validation, and testing, respectively. |
| Hardware Specification | Yes | We used 8 NVIDIA V100 32 Gi B GPUs for conducting the experiments. |
| Software Dependencies | No | The paper mentions software components like MMDetection3D, Center Point, DINOv2, and ViT-L/14, but does not provide specific version numbers for these software dependencies, only their names and general configurations. |
| Experiment Setup | Yes | More specifically, we use Center Point with pillars of 0.2 m as voxel encoder, do not use test time augmentation, and train for 20 epochs with a batch size of 4. All class-agnostic experiments are done without class-balanced grouping and sampling (CBGS) [34], while we do use CBGS for multi-class experiments to improve the performance of tail classes. |