RECORD
M$^2$DAR: Multi-View Multi-Scale Driver Action Recognition with Vision Transformer
Computer Vision and Pattern Recognition · Submitted 13 May 2023
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim Extensive experiments conducted on the 7th AI City Challenge Track 3 dataset demonstrate the effectiveness of our approach, where we achieved an overlap score of 0.5921 on the A2 test set.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
Ensuring traffic safety and preventing accidents is a critical goal in daily driving, where the advancement of computer vision technologies can be leveraged to achieve this goal. In this paper, we present a multi-view, multi-scale framework for naturalistic driving action recognition and localization in untrimmed videos, namely M$^2$DAR, with a particular focus on detecting distracted driving behaviors. Our system features a weight-sharing, multi-scale Transformer-based action recognition network that learns robust hierarchical representations. Furthermore, we propose a new election algorithm consisting of aggregation, filtering, merging, and selection processes to refine the preliminary results from the action recognition module across multiple views. Extensive experiments conducted on the 7th AI City Challenge Track 3 dataset demonstrate the effectiveness of our approach, where we achieved an overlap score of 0.5921 on the A2 test set. Our source code is available at \url{https://github.com/PurdueDigitalTwin/M2DAR}.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | Ensuring traffic safety and preventing accidents is a critical goal in daily driving, where the advancement of computer vision technologies can be leveraged to achieve this goal. | background 1.00 | 0.04 | 0.02 | asserted 0.95 |
| s1 | In this paper, we present a multi-view, multi-scale framework for naturalistic driving action recognition and localization in untrimmed videos, namely M$^2$DAR, with a particular focus on detecting distracted driving behaviors. | contribution 1.00 | 0.86 | 0.08 | asserted 0.97 |
| s2 | Our system features a weight-sharing, multi-scale Transformer-based action recognition network that learns robust hierarchical representations. | method 0.95 | 0.86 | 0.05 | asserted 1.00 |
| s3 | Furthermore, we propose a new election algorithm consisting of aggregation, filtering, merging, and selection processes to refine the preliminary results from the action recognition module across multiple views. | method 0.68 | 0.94 | 0.03 | asserted 0.98 |
| s4 | Extensive experiments conducted on the 7th AI City Challenge Track 3 dataset demonstrate the effectiveness of our approach, where we achieved an overlap score of 0.5921 on the A2 test set. | result 1.00 | 0.98 | 0.58 | asserted 1.00 |
| s5 | Our source code is available at \url{https://github.com/PurdueDigitalTwin/M2DAR}. | contribution 0.68 | 0.50 | 0.02 | asserted 0.99 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
Ensuring traffic safety and preventing accidents is a critical goal in daily driving, where the advancement of computer vision technologies can be leveraged to achieve this goal. In this paper, we present a multi-view, multi-scale framework for naturalistic driving action recognition and localization in untrimmed videos, namely M$^2$DAR, with a particular focus on detecting distracted driving behaviors. Our system features a weight-sharing, multi-scale Transformer-based action recognition network that learns robust hierarchical representations. Furthermore, we propose a new election algorithm consisting of aggregation, filtering, merging, and selection processes to refine the preliminary results from the action recognition module across multiple views. Extensive experiments conducted on the 7th AI City Challenge Track 3 dataset demonstrate the effectiveness of our approach, where we achieved an overlap score of 0.5921 on the A2 test set. Our source code is available at \url{https://github.com/PurdueDigitalTwin/M2DAR}.
Human review
How the literature uses it
What this paper does with the arXiv papers it cites
What it cites · 5 sentences without a confident label
Uses the method or tool of 1 paper
-
We utilized a pretrained MViTv2-B model on Kinetics-700 [ref] , with a frame length of $S=16$ and a sample rate of $τ=4$.
The Kinetics Human Action Video Dataset · Implementation · 0.85
Cites as background 1 paper
-
ViT-DD proposed a semi-supervised framework to make use of driver emotion as an additional clue in recognizing driver behaviors [ref] .
ViT-DD: Multi-Task Vision Transformer for Semi-Supervised Driver Distraction Detection · Related Work · 0.88
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.