Learning Semantic Segmentation from Multiple Datasets with Label Shifts

Kim, Dong-Wan; Tsai, Yi-Hsuan; Suh, Young-Ger; Faraki, Masoud; Garg, Sparsh; Chandraker, Manmohan; Han, Bohyung

doi:10.1007/978-3-031-19815-1_2

Cited by 8 publications

(3 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Transfer learning allows the reuse of pre-trained neural networks to adapt them to solve other tasks. Using a neural network pre-trained by a source dataset as a starting point, we can reduce the amount of data and annotations in the target dataset required to train a neural network to solve the target task [9]. However, using this approach is not trivial, as it will only work in situations where the domains of both image datasets are relatively related.…”

Section: Introductionmentioning

confidence: 99%

OffRoadSynth Open Dataset for Semantic Segmentation using Synthetic-Data-Based Weight Initialization for Autonomous UGV in Off-Road Environments

Małek,

Dybała,

Kordecki

et al. 2024

J Intell Robot Syst

View full text Add to dashboard Cite

This article concerns the issue of image semantic segmentation for the machine vision system of an autonomous Unmanned Ground Vehicle (UGV) moving in an off-road environment. Determining the meaning (semantics) of the areas visible in the recorded image provides a complete understanding of the scene surrounding the autonomous vehicle. It is crucial for the correct determination of a passable route. Nowadays, semantic segmentation is generally solved using convolutional neural networks (CNN), which can take an image as input and output the segmented image. However, proper training of the neural network requires the use of large amounts of data, which becomes problematic in the situation of low availability of large, dedicated image data sets that consider various off-road situations - driving on various types of roads, surrounded by diverse vegetation and in various weather and light conditions. This study introduces a synthetic image dataset called “OffRoadSynth” to address the training data scarcity for off-road scenarios. It has been shown that pre-training the neural network on this synthetic dataset improves image segmentation accuracy compared to other methods, such as random network weight initialization or using larger, generic datasets. Results suggest that using a smaller but domain-dedicated set of synthetic images to initialize network weights for training on the target real-world dataset may be an effective approach to improving semantic segmentation results of images, including those from off-road environments.

show abstract

Section: Introductionmentioning

confidence: 99%

OffRoadSynth Open Dataset for Semantic Segmentation using Synthetic-Data-Based Weight Initialization for Autonomous UGV in Off-Road Environments

Małek,

Dybała,

Kordecki

et al. 2024

J Intell Robot Syst

View full text Add to dashboard Cite

show abstract

“…Leveraging data samples from different sources for training has been proven effective in enhancing robustness and generalizability [74]. Various approaches have been proposed to merge image datasets for object detection [15,60,62,101,127,128], image segmentation [31,44,45,55,126], depth estimation [14,84], etc. Due to large domain gaps, the image-based methods are often hard to be transferred to 3D.…”

Section: Related Workmentioning

confidence: 99%

“…In this work, we define a novel paradigm towards leveraging LiDAR point clouds from different datasets to tame a single set of parameters for multi-task LiDAR segmentation. Sibling to image segmentation communities [45,55,126], we call this paradigm universal LiDAR segmentation. The ultimate goal of such a synergistic way of learning is to build a powerful segmentation model that can absorb rich cross-domain knowledge and, in return, achieve strong resilience and generalizability for practical usage.…”

Section: Introductionmentioning

confidence: 99%

Exploring metadiscourse in university lectures in Hong Kong

Liu¹

View full text Add to dashboard Cite

Federated Graph Neural Network (FedGNN) has recently emerged as a rapidly growing research topic, as it integrates the strengths of graph neural networks and federated learning to enable advanced machine learning applications without direct access to sensitive data. Despite its advantages, the distributed nature of FedGNN introduces additional vulnerabilities, particularly backdoor attacks stemming from malicious participants. Although graph backdoor attacks have been explored, the compounded complexity introduced by the combination of GNNs and federated learning has hindered a comprehensive understanding of these attacks, as existing research lacks extensive benchmark coverage and in-depth analysis of critical factors. To address these limitations, we propose Bkd-FedGNN, a benchmark for backdoor attacks on FedGNN. Specifically, Bkd-FedGNN decomposes the graph backdoor attack into trigger generation and injection steps, and extending the attack to the node-level federated setting, resulting in a unified framework that covers both node-level and graph-level classification tasks. Moreover, we thoroughly investigate the impact of multiple critical factors in backdoor attacks on FedGNN. These factors are categorized into global-level and local-level factors, including data distribution, the number of malicious attackers, attack time, overlapping rate, trigger size, trigger type, trigger position, and poisoning rate. Finally, we conduct comprehensive evaluations on 13 benchmark datasets and 13 critical factors, comprising 1,725 experimental configurations for node-level and graphlevel tasks from six domains. These experiments encompass over 8,000 individual tests, allowing us to provide a thorough evaluation and insightful observations that advance our understanding of backdoor attacks on FedGNN. The Bkd-FedGNN benchmark is publicly available at https://github.com/usail-hkust/ BkdFedGCN.Preprint. Under review.

show abstract