This repository contains the dataset relevant for the following publication.
Sreyasi Nag Chowdhury, Ruwan Wickramarachchi, Mohamed H. Gad-Elrab, Daria Stepanova, and Cory Henson. Towards Leveraging Commonsense Knowledge for Autonomous Driving. ISWC 2021 - Demos & Posters track. [Text]
More specifically, the dataset stores a collection of commonsense knowledge assertions relevant for the autonomous driving (AD) domain, which can be used either independently or in combination with pandaset (an open-source dataset for urban AD situations).
Please cite the above paper if using the dataset.
The repository stores the following data:
(1) pandaset.ttl - A knowledge graph [1] constructed relying on
(2) pandaset_ad_csk.ttl - An extension of pandaset.ttl with commonsense relations from:
(3) link_prediction_models - A folder with the link prediction models TransE [2] and HolE [3] pretrained using Ampligraph 1.4 on (1) and (2). To load the models run
from ampligraph.utils import restore_model
model = restore_model('<model_name>.pkl')
(4) explainable_clustering - A folder with the explainable clustering results, i.e., the output of the method from [4] on the following input:
- pandaset_no_literals.tsv: The KG (1) without literals in the tsv format
- pandaset_ad_csk_no_literals.tsv: The KG (2) without literals in the tsv format
- target_scene_entities.txt: Scenes to be clustered
[1] Ruwan Wickramarachchi, Cory Henson, and Amit Sheth. Knowledge-infused Learning for Entity Prediction in Driving Scenes. Frontiers in Big Data 4 (2021): 759110.
[2] Antoine Bordes, Nicolas Usunier, Alberto García-Durán, Jason Weston, Oksana Yakhnenko: Translating Embeddings for Modeling Multi-relational Data. NIPS 2013: 2787-2795
[3] Maximilian Nickel, Lorenzo Rosasco, Tomaso A. Poggio: Holographic Embeddings of Knowledge Graphs. AAAI 2016: 1955-1961
[4] Mohamed H. Gad-Elrab, Daria Stepanova, Trung-Kien Tran, Heike Adel, Gerhard Weikum: ExCut: Explainable Embedding-Based Clustering over Knowledge Graphs. ISWC (1) 2020: 218-237
This dataset is licensed under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC-BY-SA-4.0).