Leveraging Privileged Information to Limit Distraction in End-to-End Lane Following

Communications in Computer and Information Science

Frank

2022

Free space estimation is an important problem for autonomous robot navigation. Traditional camera-based approaches rely on pixel-wise ground truth annotations to train a segmentation model. To cover the wide variety of environments and lighting conditions encountered on roads, training supervised models requires large datasets. This makes the annotation cost prohibitively high. In this work, we propose a novel approach for obtaining free space estimates from images taken with a single road-facing camera. We rely on a technique that generates weak free space labels without any supervision, which are then used as ground truth to train a segmentation model for free space estimation. We study the impact of different data augmentation techniques on the performances of free space predictions, and propose to use a recursive training strategy. Our results are benchmarked using the Cityscapes dataset and improve over comparable published work across all evaluation metrics. Our best model reaches 83.64% IoU (+2.3%), 91.75% Precision (+2.4%) and 91.29% Recall (+0.4%). These results correspond to 88.8% of the IoU, 94.3% of the Precision and 93.1% of the Recall obtained by an equivalent fully-supervised baseline, while using no ground truth annotation. Our code and models are freely available online.

Section: Training Strategies For Weakly-supervised Segmentationmentioning

confidence: 99%

Refining Weakly-Supervised Free Space Estimation Through Data Augmentation and Recursive Training

Communications in Computer and Information Science

Frank

2022

“…In the first and third images, specific road characteristics such as road pavement and a pedestrian crossing are likely throwing the model off. This could be prevented by leveraging a visualization of the network focus at training time to encouraging it to focus on the road itself [19]. In the second image, the model attempts to take a right turn while the human demonstrator only decided to leave the roundabout at the next exit.…”

Section: A Learning Steering Angles From a Single Cameramentioning

confidence: 99%

RoboBus: A Diverse and Cross-Border Public Transport Dataset

Varisteas

Frank

2021 IEEE International Conference on Pervasive Computing and Communications Workshops and Other Affiliated Events (PerCom Work

2021

Self Cite

Academic datasets are an important source of information to validate and benchmark novel research concepts. In this paper we present RoboBus, a dataset recorded with a commercial bus on a cross-border public transport route between Luxembourg and France. The dataset contains approximately 8 hours of driving data divided into 15 trips that have been recorded over 4 days. It includes about 1.7 million anonymized images captured by two road-facing cameras, GNSS traces, data from a 9-axis IMU, and information directly retrieved from the CAN interface of the vehicle including speed, steering angle and position of the accelerator/brake pedals. We use an end-to-end autonomous driving approach that relies on imitation learning as use case example for the dataset.

Weakly-Supervised Free Space Estimation through Stochastic Co-Teaching

2022 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW)

Parera

Hundt³

et al. 2022

Self Cite

Free space estimation is an important problem for autonomous robot navigation. Traditional camera-based approaches train a segmentation model using an annotated dataset. The training data needs to capture the wide variety of environments and weather conditions encountered at runtime, making the annotation cost prohibitively high. In this work, we propose a novel approach for obtaining free space estimates from images taken with a single roadfacing camera. We rely on a technique that generates weak free space labels without any supervision, which are then used as ground truth to train a segmentation model for free space estimation. Our work differs from prior attempts by explicitly taking label noise into account through the use of Co-Teaching. Since Co-Teaching has traditionally been investigated in classification tasks, we adapt it for segmentation and examine how its parameters affect performances in our experiments. In addition, we propose Stochastic Co-Teaching, which is a novel method to select clean samples that leads to enhanced results. We achieve an IoU of 82.6%, a Precision of 90.9%, and a Recall of 90.3%. Our best model reaches 87% of the IoU, 93% of the Precision, and 93% of the Recall of the equivalent fully-supervised baseline while using no human annotations. To the best of our knowledge, this work is the first to use Co-Teaching to train a free space segmentation model under explicit label noise. Our implementation and models are freely available online.