United States Department of Transportation
i
User Trajectory Estimation from Visual Features: Development of Inner-Ensemble Averaging (IEA) for Deep Learning
-
2019-09-01
[PDF-1.43 MB]
Details:
-
Creators:
-
Corporate Creators:
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Publication/ Report Number:
-
Resource Type:
-
Geographical Coverage:
-
Edition:Interim Report
-
Corporate Publisher:
-
Abstract:Detecting human behavior (for example grabbing the top of a steering wheel before a turn, or looking over the shoulder for a bicyclist) for the purposes of building a model to predict future vehicle trajectories (to avoid collisions) is very difficult. These gestures are probably very strong predictors for forecasting trajectories over a short time horizon. However, at present there are not any practical, scalable traffic safety systems that consider human body cues to predict vehicle trajectories. Deep learning (DL) offers a potential way of inferring future user trajectories, through the analysis of user trajectory datasets, and the use of neural networks for detecting users positions and relevant user features. However, most deep learning models fail at capturing the complexity of the problem. One of the main issues of currently used models is their lack of generalization on larger datasets (overfit). In this project, the objective is to develop a new class of deep models based on ensemble averaging (EA). Ensemble averaging allows the results of the model to be more robust to changes in the data, and therefore leads to models that have better predictive power. The Inner Ensemble Average method is introduced. The method is validated in benchmark image recognition datasets that are standard in the Machine Learning community. This method will then be applied to trajectory detection and user position inference in future work.
-
Format:
-
Funding:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
