Semantic object classes in video: A high-definition ground truth database

By Gabriel J. Brostow, Julien Fauqueur, Roberto Cipolla. 2009

1,529 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Evaluation & Benchmarks

Citations in the corpus

Microsoft COCO: Common Objects in Context (2014, background): “a bounding box [2] or with a precise pixel-level segmentation [14], [15], [16].”
The Cityscapes Dataset for Semantic Urban Scene Understanding (2016, methodology): “Also in this area, research progress can be heavily linked to the existence of datasets such as the KITTI Vision Benchmark Suite [19], CamVid [7], Leuven [35], and Daimler Urban Segmentation [61] datasets.”
COCO-Stuff: Thing and Stuff Classes in Context (2016, methodology): “In particular, MSRC 21 [46], KITTI [23], CamVid [7], SIFT Flow [36] and NYUD [47] all have less than 5,000 images (Table 1).”
Occluded Video Instance Segmentation: A Benchmark (2021, methodology): “CamVid [6] dataset contains 4 videos and the authors annotate one frame every 30 frames, obtaining 800 annotated frames finally.”
SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation (2015, methodology): “A number of road scene datasets are available for semantic parsing [21], [25], [59], [60].”
Learning to Prune Filters in Convolutional Neural Networks (2018, methodology): “For the segmentation task, we test the FCN-32s network [24] on Pascal VOC dataset [8] and the SegNet network [2] on the CamVid dataset [4].”
Driving Policy Transfer via Modularity and Abstraction (2018, methodology): “For training, we use standard public datasets – CamVid [5], Cityscapes [9], Berkeley Driving [52], and a combination of all these – as well as our small dataset and data collected in simulation.”
Playing for Data: Ground Truth from Computer Games (2016, methodology): “To evaluate the utility of using game data for training semantic segmentation systems, we used label definitions compatible with other datasets for urban scene understanding [8, 11, 13, 52].”
Bayesian SegNet: Model Uncertainty in Deep Convolutional Encoder-Decoder Architectures for Scene Understanding (2015, methodology): “4 shows segmentations and model uncertainty results from Bayesian SegNet on CamVid Road Scenes [3].”
Deep Layer Aggregation (2017, methodology): “We further transfer the learned network to semantic image segmentation for CamVid (Brostow et al. 2009) and Cityscapes (Cordts et al. 2016), and beat the previous best method by 4 points with fewer parameters on CamVid dataset.”
Universal Adversarial Perturbations Against Semantic Image Segmentation (2017, methodology): “Generalizability We have tested the effect of the universal perturbation generated for Cityscapes on CamVid [1] (without any fine-tuning on CamVid).”
MSeg: A Composite Dataset for Multi-Domain Semantic Segmentation (2020, background): “Training & Validation COCO [19] Everyday objects 123,287 + COCO STUFF [4] ADE20K [46] Everyday objects 22,210 MAPILLARY [25] Driving (Worldwide) 20,000 IDD [40] Driving (India) 7,974 BDD [43] Driving (United States) 8,000 CITYSCAPES [7] Driving (Germany) 3,475 SUN RGBD [36] Indoor 5,285 Test PASCAL ”
Survey on semantic segmentation using deep learning techniques (2019, methodology): “The automotive datasets includes; CamVid dataset [180] which is considered as the ﬁrst with semantically annotated videos, Daimler Urban Segmentation [181], CityScapes [182], Mapillary Vistas [183] and the most recent Apolloscape-Scene parsing [184] which focuses on semantic understanding of urban…”
Procedural Modeling and Physically Based Rendering for Synthetic Data Generation in Automotive Applications (2017, background): “CamVid [3], Cityscapes [5], the KITTI dataset [9], or the Mapillary Vistas Dataset2, which is both time consuming and complex to orchestrate.”
Curriculum Domain Adaptation for Semantic Segmentation of Urban Scenes (2017, background): “According to [47], it took about 60 minutes to manually segment each image in [8] and about 90 minutes for each in [12].”