Fig 1 - uploaded by Joshua Marshall
Content may be subject to copyright.
![Imbalance in the Freiburg Forest dataset [6]: the Tree and Obstacle classes are imbalanced; and obstacles appear in more images than trees, but have a smaller pixel count because they are physically smaller, showing spatial imbalance.](profile/Joshua-Marshall-2/publication/353020613/figure/fig1/AS:1087580925894657@1636310907060/mbalance-in-the-Freiburg-Forest-dataset-6-the-Tree-and-Obstacle-classes-are.png)
Imbalance in the Freiburg Forest dataset [6]: the Tree and Obstacle classes are imbalanced; and obstacles appear in more images than trees, but have a smaller pixel count because they are physically smaller, showing spatial imbalance.
![Fig. 1. Imbalance in the Freiburg Forest dataset [6]: the Tree and...](profile/Joshua-Marshall-2/publication/353020613/figure/fig1/AS:1087580925894657@1636310907060/mbalance-in-the-Freiburg-Forest-dataset-6-the-Tree-and-Obstacle-classes-are_Q320.jpg)
Contexts in source publication
Context 1
... in remote off-road environments. Consequently, in off-road datasets, an obstacle class is often defined comprising an eclectic range of non-unique natural and man-made objects such as rocks of varying sizes and shapes, fences, power pylons, etc. A further complication is the natural class and spatial imbalance in the datasets; e.g., as shown in Fig. 1 for the Freiburg Forest dataset [6] in which obstacles are represented in more images than trees, but occupy fewer pixels showing spatial ...
Context 2
... Traversable grass, Grass and Vegetation; representative of three levels of difficulty of traversal. It also includes a Water class. The SOOR dataset consists of 257 densely labelled images of size 768×384 pixels taken on three separate days in July in southern Ontario, Canada. Both datasets exhibit natural class and sample imbalance as shown in Figs. 1 and ...
Context 3
... from the Freiburg Forest dataset are shown in Table II. The most difficult class in this dataset is the Objects class ( Fig. 1) for the same reasons as in the SOOR dataset. Merging the Tree and Vegetation classes, improves results particularly in the Object class [6]. But we considered the more challenging case by training the CNN with separate Tree and Vegetation classes. However, the Tree class is not represented in their test set, hence the overall ...
Similar publications
La présente étude a pour but d’explorer le processus d’élaboration et de validation de contenu de tâches d’évaluation de la production orale en allemand langue étrangère chez des enfants de Suisse romande.
Les tâches, conçues pour évaluer les objectifs correspondants du plan d’études romand (PER), ont la particularité d’être des tâches numériques q...
The problem of predicting driver attention from the driving perspective is gaining the increasing research focuses due to its remarkable significance for autonomous driving and assisted driving systems. Driving experience is extremely important for driver attention prediction, a skilled driver is able to effortlessly predict oncoming danger (before...
There are fifty second-order ordinary differential equations of Painlev\'e type in the literature. This paper investigates the Mei symmetries corresponding to the Lagrangian of these equations. This classifies these equations by the admitted Mei symmetries, along with the Mei invariants and their respective gauge functions. The existence of Mei sym...
Ttraditional safety engineering is coming to a turning point moving from deterministic, non-evolving systems operating in well-defined contexts to increasingly autonomous and learning-enabled AI systems which are acting in largely unpredictable operating contexts. We outline some of underlying challenges of safe AI and suggest a rigorous engineerin...
Citations
... Since we target off-road applications, we use a base dataset recorded in off-road environments with a diverse range of objects and terrain types. From the online available datasets [13]- [15] we choose Rellis3D [14], which is collected in an off-road environment with 6,235 annotated images. The classes in this dataset are Concrete, Asphalt, Gravel, Grass, Dirt, Sand, Rock, Rock Bed, Water, Bushes, Tall Vegetation, Trees, Poles, Logs, Void, Sky, Sign. ...
For navigation of robots, image segmentation is an important component to determining a terrain's traversability. For safe and efficient navigation, it is key to assess the uncertainty of the predicted segments. Current uncertainty estimation methods are limited to a specific choice of model architecture, are costly in terms of training time, require large memory for inference (ensembles), or involve complex model architectures (energy-based, hyperbolic, masking). In this paper, we propose a simple, light-weight module that can be connected to any pretrained image segmentation model, regardless of its architecture, with marginal additional computation cost because it reuses the model's backbone. Our module is based on maximum separation of the segmentation classes by respective prototype vectors. This optimizes the probability that out-of-distribution segments are projected in between the prototype vectors. The uncertainty value in the classification label is obtained from the distance to the nearest prototype. We demonstrate the effectiveness of our module for terrain segmentation.
... However, the efforts focused on off-road navigation are still less extensive compared to on-road, due to the lack of off-road consistent and structured datasets. The main datasets available for training a semantic segmentation model for off-road autonomous navigation are: RUGD [1], RELLIS-3D [4], DeepScene-Freiburg-Forest [5], YCOR [6], SOOR [7] and very recently ORFD [8]. These datasets demonstrate certain differences with respect to the number of classes present in the data as well as variation on image resolution and aspect ratio. ...
... The architecture was trained with the RUGD dataset and the RELLIS-3D dataset. In [7], the authors combine DeepScene-Freiburg-Forest dataset and SOOR dataset for training. They fuse RGB and texture descriptors (i.e. ...
Off-road autonomous unmanned ground vehicles (UGVs) are being developed for military and commercial use to deliver crucial supplies in remote locations, help with mapping and surveillance, and to assist war-fighters in contested environments. Due to complexity of the off-road environments and variability in terrain, lighting conditions, diurnal and seasonal changes, the models used to perceive the environment must handle a lot of input variability. Current datasets used to train perception models for off-road autonomous navigation lack of diversity in seasons, locations, semantic classes, as well as time of day. We test the hypothesis that model trained on a single dataset may not generalize to other off-road navigation datasets and new locations due to the input distribution drift. Additionally, we investigate how to combine multiple datasets to train a semantic segmentation-based environment perception model and we show that training the model to capture uncertainty could improve the model performance by a significant margin. We extend the Masksembles approach for uncertainty quantification to the semantic segmentation task and compare it with Monte Carlo Dropout and standard baselines. Finally, we test the approach against data collected from a UGV platform in a new testing environment. We show that the developed perception model with uncertainty quantification can be feasibly deployed on an UGV to support online perception and navigation tasks.
Off‐road autonomous vehicles (OAVs) are becoming increasingly popular for navigating challenging environments in agriculture, military, and exploration applications. These vehicles face unique challenges, such as unpredictable terrain, dynamic obstacles, and varying environmental conditions. Therefore, it is essential to have an efficient terrain classification system to ensure safe and efficient operation of OAVs. This paper provides an overview of recent advances and emerging trends in off‐road terrain classification methods. Through a comprehensive literature review, this study explores the use of sensor modalities and techniques that leverage both appearance and geometry of the terrain for classification tasks. The study discusses learning‐based approaches, particularly deep learning, and highlights the integration of multiple sensor modalities through hybrid multimodal techniques. Finally, this study reviews the available off‐road datasets and explores the use cases and applications of terrain classification across various autonomous domains. Given the rapid advancements in terrain classification, this paper organizes and surveys to provide a comprehensive overview. By offering a structured review of the current landscape, this paper significantly enhances our understanding of terrain classification in unstructured environments, while also highlighting important areas for future research, particularly in deep‐learning‐based advancements.
A recently developed application of computer vision is pathfinding in self-driving cars. Semantic scene understanding and semantic segmentation, as subfields of computer vision, are widely used in autonomous driving. Semantic segmentation for pathfinding uses deep learning methods and various large sample datasets to train a proper model. Due to the importance of this task, accurate and robust models should be trained to perform properly in different lighting and weather conditions and in the presence of noisy input data. In this paper, we propose a novel learning method for semantic segmentation called layer-wise training and evaluate it on a light efficient structure called an efficient neural network (ENet). The results of the proposed learning method are compared with the classic learning approaches, including mIoU performance, network robustness to noise, and the possibility of reducing the size of the structure on two RGB image datasets on the road (CamVid) and off-road (Freiburg Forest) paths. Using this method partially eliminates the need for Transfer Learning. It also improves network performance when input is noisy.
