In traditional semantic segmentation, knowing about all existing classes is essential to yield effective results with the majority of existing approaches. However, these methods trained in a Closed Set of classes fail when new classes are found in the test phase, not being able to recognize that an unseen class has been fed. This means that they are not suitable for Open Set scenarios, which are very common in real-world computer vision and remote sensing applications. In this paper, we discuss the limitations of Closed Set segmentation and propose two fully convolutional approaches to effectively address Open Set semantic segmentation: OpenFCN and OpenPCS. OpenFCN is based on the well-known OpenMax algorithm, configuring a new application of this approach in segmentation settings. Open-PCS is a fully novel approach based on feature-space from DNN activations that serve as features for computing PCA and multi-variate gaussian likelihood in a lower dimensional space. In addition to OpenPCS and aiming to reduce the RAM memory requirements of the methodology, we also propose a slight variation of the method (OpenIPCS) that uses an iteractive version of PCA able to be trained in small batches. Experiments were conducted on the well-known ISPRS Vaihingen/Potsdam and the 2018 IEEE GRSS Data Fusion Challenge datasets. OpenFCN showed little-to-no improvement when compared to the simpler and much more time efficient SoftMax thresholding, while being some orders of magnitude slower. OpenPCS achieved promising results in almost all experiments by overcoming both OpenFCN and SoftMax thresholding. OpenPCS is also a reasonable compromise between the runtime performances of the extremely fast SoftMax thresholding and the extremely
It is undeniable that aerial/satellite images can provide useful information for a large variety of tasks. But, since these images are always looking from above, some applications can benefit from complementary information provided by other perspective views of the scene, such as ground-level images. Despite a large number of public repositories for both georeferenced photographs and aerial images, there is a lack of benchmark datasets that allow the development of approaches that exploit the benefits and complementarity of aerial/ground imagery. In this paper, we present two new publicly available datasets named AiRound and CV-BrCT. The first one contains triplets of images from the same geographic coordinate with different perspectives of view extracted from various places around the world. Each triplet is composed of an aerial RGB image, a ground-level perspective image, and a Sentinel-2 sample. The second dataset contains pairs of aerial and street-level images extracted from southeast Brazil. We design an extensive set of experiments concerning multi-view scene classification, using early and late fusion. Such experiments were conducted to show that image classification can be enhanced using multi-view data.
Fault picking is a critical, but human-labor-intensive component of seismic interpretation. In a bid to improve fault imaging in seismic data, we have applied a directional Laplacian of a Gaussian operator to sharpen fault features within a coherence volume. We computed an M × M matrix of the second moment tensor distanceweighted coherence values that fell within a 3D analysis window about each voxel. The eigenvectors of this matrix defined the orientation of planar discontinuities, whereas the corresponding eigenvalues determined whether these discontinuities were significant. The eigenvectors, which quantified the fault dip magnitude and dip azimuth, defined a natural coordinate system for smoothing of the planar discontinuity. We rotated the data to the new coordinate system and applied the sharpening operator. By comparing the vector dip of the discontinuity to the vector dip of the reflectors, we could apply a filter to either suppress or enhance discontinuities associated with unconformities or low-signal-to-noise-ratio shale-on-shale reflectors. We have revealed the value and robustness of the technique by application to two 3D data volumes from offshore New Zealand, which exhibited polygonal faulting, shale dewatering, and mass transport complexes.
Improving the accuracy and completeness of subtle discontinuities in noisy seismic data is useful for mapping faults, fractures, unconformities, and stratigraphic edges. We have developed a workflow to improve the quality of coherence attributes. First, we apply principal component structure-oriented filtering to reject random noise and sharpen the lateral edges of seismic amplitude data. Next, we compute eigenstructure coherence, which highlights the stratigraphic and structural discontinuities. We apply a Laplacian of a Gaussian filter to the coherence attribute that sharpens the steeply dipping faults, attenuates the stratigraphic features parallel to the seismic reflectors, and skeletonizes the unconformity features subparallel to the reflectors. Finally, we skeletonize the filtered coherence attribute along with the fault plane. The filtered and skeletonized seismic coherence attribute highlights the geologic discontinuities more clearly and precisely. These discontinuous features can be color coded by their dipping orientation or as a suite of independent, azimuthally limited volumes, providing the interpreter a means of isolating fault sets that are either problematic or especially productive. We validate the effectiveness of our workflow by applying it to seismic surveys acquired from the Gulf of Mexico, USA, and the Great South Basin, New Zealand. The skeletonized result rejects noise and enhances discontinuities seen in the vertical and lateral directions. The corendering of the “fault” azimuth and the fault-dip magnitude exhibits the strengths of the discontinuities and their orientation. Finally, we compared our workflow with the results generated from the swarm intelligence and found our method to be better at tracking short faults and stratigraphic discontinuities.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.