Explainable action prediction through self-supervision on scene graphs
This work explores scene graphs as a distilled representation of high-level information for autonomous driving, applied to future driver-action prediction. Given the scarcity and strong imbalance of data samples, we propose a selfsupervision pipeline to infer representative and well-separated embedd...
| Main Authors: | , , , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
IEEE
2023
|
| _version_ | 1826310579647676416 |
|---|---|
| author | Kochakarn, P De Martini, D Omeiza, D Kunze, L |
| author_facet | Kochakarn, P De Martini, D Omeiza, D Kunze, L |
| author_sort | Kochakarn, P |
| collection | OXFORD |
| description | This work explores scene graphs as a distilled representation of high-level information for autonomous driving,
applied to future driver-action prediction. Given the scarcity
and strong imbalance of data samples, we propose a selfsupervision pipeline to infer representative and well-separated
embeddings. Key aspects are interpretability and explainability;
as such, we embed in our architecture attention mechanisms
that can create spatial and temporal heatmaps on the scene
graphs. We evaluate our system on the ROAD dataset against
a fully-supervised approach, showing the superiority of our
training regime. |
| first_indexed | 2024-03-07T07:54:03Z |
| format | Conference item |
| id | oxford-uuid:1a6a84e6-277d-43eb-b11e-69529283b99b |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T07:54:03Z |
| publishDate | 2023 |
| publisher | IEEE |
| record_format | dspace |
| spelling | oxford-uuid:1a6a84e6-277d-43eb-b11e-69529283b99b2023-08-09T09:46:45ZExplainable action prediction through self-supervision on scene graphsConference itemhttp://purl.org/coar/resource_type/c_5794uuid:1a6a84e6-277d-43eb-b11e-69529283b99bEnglishSymplectic ElementsIEEE2023Kochakarn, PDe Martini, DOmeiza, DKunze, LThis work explores scene graphs as a distilled representation of high-level information for autonomous driving, applied to future driver-action prediction. Given the scarcity and strong imbalance of data samples, we propose a selfsupervision pipeline to infer representative and well-separated embeddings. Key aspects are interpretability and explainability; as such, we embed in our architecture attention mechanisms that can create spatial and temporal heatmaps on the scene graphs. We evaluate our system on the ROAD dataset against a fully-supervised approach, showing the superiority of our training regime. |
| spellingShingle | Kochakarn, P De Martini, D Omeiza, D Kunze, L Explainable action prediction through self-supervision on scene graphs |
| title | Explainable action prediction through self-supervision on scene graphs |
| title_full | Explainable action prediction through self-supervision on scene graphs |
| title_fullStr | Explainable action prediction through self-supervision on scene graphs |
| title_full_unstemmed | Explainable action prediction through self-supervision on scene graphs |
| title_short | Explainable action prediction through self-supervision on scene graphs |
| title_sort | explainable action prediction through self supervision on scene graphs |
| work_keys_str_mv | AT kochakarnp explainableactionpredictionthroughselfsupervisiononscenegraphs AT demartinid explainableactionpredictionthroughselfsupervisiononscenegraphs AT omeizad explainableactionpredictionthroughselfsupervisiononscenegraphs AT kunzel explainableactionpredictionthroughselfsupervisiononscenegraphs |