Accurate and comprehensive semantic segmentation of Bird's Eye View (BEV) is essential for ensuring safe and proactive navigation in autonomous driving. Although cooperative perception has exceeded the detection capabilities of single-agent systems, prevalent camera-based algorithms in cooperative perception neglect valuable information derived from historical observations. This limitation becomes critical during sensor failures or communication issues as cooperative perception reverts to single-agent perception, leading to degraded performance and incomplete BEV segmentation maps. This paper introduces TempCoBEV, a temporal module designed to incorporate historical cues into current observations, thereby improving the quality and reliability of BEV map segmentations. We propose an importance-guided attention architecture to effectively integrate temporal information that prioritizes relevant properties for BEV map segmentation. TempCoBEV is an independent temporal module that seamlessly integrates into state-of-the-art camera-based cooperative perception models. We demonstrate through extensive experiments on the OPV2V dataset that TempCoBEV performs better than non-temporal models in predicting current and future BEV map segmentations, particularly in scenarios involving communication failures. We show the efficacy of TempCoBEV and its capability to integrate historical cues into the current BEV map, improving predictions under optimal communication conditions by up to 2% and under communication failures by up to 19%. The code will be published on GitHub.
A novel structure learning algorithm for Bayesian Networks based on a Physarum Learner is presented. The length of the connections within an initially fully connected Physarum-Maze is taken as the inverse Pearson correlation coefficient between the connected nodes. The Physarum Learner then estimates the shortest indirect paths between each pair of nodes. In each iteration, a score of the surviving edges is incremented. Finally, the highest scored connections are combined to form a Bayesian Network. The novel Physarum Learner method is evaluated with different configurations and compared to the LAGD Hill Climber showing comparable performance with respect to quality of training results and increased time efficiency for large data sets.
AImotion Challenge Results: a Framework for AirSim Autonomous Vehicles and Motion Replication
(2023)
Efficient Noninvasive FHB Estimation using RGB Images from a Novel Multiyear, Multirater Dataset
(2023)
Fusarium head blight (FHB) is one of the most prevalent wheat diseases, causing substantial yield losses and health risks. Efficient phenotyping of FHB is crucial for accelerating resistance breeding, but currently used methods are time-consuming and expensive. The present article suggests a noninvasive classification model for FHB severity estimation using red–green–blue (RGB) images, without requiring extensive preprocessing. The model accepts images taken from consumer-grade, low-cost RGB cameras and classifies the FHB severity into 6 ordinal levels. In addition, we introduce a novel dataset consisting of around 3,000 images from 3 different years (2020, 2021, and 2022) and 2 FHB severity assessments per image from independent raters. We used a pretrained EfficientNet (size b0), redesigned as a regression model. The results demonstrate that the interrater reliability (Cohen’s kappa, κ) is substantially lower than the achieved individual network-to-rater results, e.g., 0.68 and 0.76 for the data captured in 2020, respectively. The model shows a generalization effect when trained with data from multiple years and tested on data from an independent year. Thus, using the images from 2020 and 2021 for training and 2022 for testing, we improved the Fw1 score by 0.14, the accuracy by 0.11, κ by 0.12, and reduced the root mean squared error by 0.5 compared to the best network trained only on a single year’s data. The proposed lightweight model and methods could be deployed on mobile devices to automatically and objectively assess FHB severity with images from low-cost RGB cameras. The source code and the dataset are available at https://github.com/cvims/FHB_classification.
In this paper, we evaluate the training of GAN for synthetic RAD image generation for four objects reflected by Frequency Modulated Continuous Wave radar: car, motorcycle, pedestrian and truck. This evaluation adds a new possibility for data augmentation when radar data labeling available is not enough. The results show that, yes, the GAN generated RAD images well, even when a specific class of the object is necessary. We also compared the scores of three GAN architectures, GAN Vanilla, CGAN, and DCGAN, in RAD synthetic imaging generation. We show that the generator can produce RAD images well enough with the results analyzed.