Inspired by recent work in machine translation and object detection, we introduce an attention based model that automatically learns to describe the content of images. We describe how we can train this model in a deterministic manner using standard backpropagation techniques and stochastically by maximizing a variational lower bound. We also show through visualization how the model is able to automatically learn to fix its gaze on salient objects while generating the corresponding words in the output sequence. We validate the use of attention with state-of-the-art performance on three benchmark datasets: Flickr9k, Flickr30k and MS COCO.

/pdf/show-attend-and-tell-neural-image-caption-generation-with-2atzmg08ji.pdf

Show, Attend and Tell: Neural Image Caption Generation with Visual Attention

Inspired by recent work in machine translation and object detection, we introduce an attention based model that automatically learns to describe the content of images. We describe how we can train this model in a deterministic manner using standard backpropagation techniques and stochastically by maximizing a variational lower bound. We also show through visualization how the model is able to automatically learn to fix its gaze on salient objects while generating the corresponding words in the output sequence. We validate the use of attention with state-of-the-art performance on three benchmark datasets: Flickr8k, Flickr30k and MS COCO.

For the first time in the International Terrestrial Reference Frame (ITRF) history, the ITRF2014 is generated with an enhanced modeling of nonlinear station motions, including seasonal (annual and semiannual) signals of station positions and postseismic deformation for sites that were subject to major earthquakes. Using the full observation history of the four space geodetic techniques (very long baseline interferometry (VLBI), satellite laser ranging (SLR), Global Navigation Satellite Systems (GNSS), and Doppler orbitography and radiopositioning integrated by satellite (DORIS)), the corresponding international services provided reprocessed time series (weekly from SLR and DORIS, daily from GNSS, and 24 h session-wise from VLBI) of station positions and daily Earth Orientation Parameters. ITRF2014 is demonstrated to be superior to past ITRF releases, as it precisely models the actual station trajectories leading to a more robust secular frame and site velocities. The ITRF2014 long-term origin coincides with the Earth system center of mass as sensed by SLR observations collected on the two LAGEOS satellites over the time span between 1993.0 and 2015.0. The estimated accuracy of the ITRF2014 origin, as reflected by the level of agreement with the ITRF2008 (both origins are defined by SLR), is at the level of less than 3 mm at epoch 2010.0 and less than 0.2 mm/yr in time evolution. The ITRF2014 scale is defined by the arithmetic average of the implicit scales of SLR and VLBI solutions as obtained by the stacking of their respective time series. The resulting scale and scale rate differences between the two solutions are 1.37 (±0.10) ppb at epoch 2010.0 and 0.02 (±0.02) ppb/yr. While the postseismic deformation models were estimated using GNSS/GPS data, the resulting parametric models at earthquake colocation sites were applied to the station position time series of the three other techniques, showing a very high level of consistency which enforces more the link between techniques within the ITRF2014 frame. The users should be aware that the postseismic deformation models are part of the ITRF2014 products, unlike the annual and semiannual signals, which were estimated internally with the only purpose of enhancing the velocity field estimation of the secular frame.

https://agupubs.onlinelibrary.wiley.com/doi/epdf/10.1002/2016JB013098

ITRF2014: A new release of the International Terrestrial Reference Frame modeling nonlinear station motions

Continuous Global Positioning System sites in southwestern British Columbia, Canada, and northwestern Washington state, USA, have been moving landward as a result of the locked state of the Cascadia subduction fault offshore In the summer of 1999, a cluster of seven sites briefly reversed their direction of motion No seismicity was associated with this event The sudden displacements are best explained by ∼2 centimeters of aseismic slip over a 50-kilometer-by-300-kilometer area on the subduction interface downdip from the seismogenic zone, a rupture equivalent to an earthquake of moment magnitude 67 This provides evidence that slip of the hotter, plastic part of the subduction interface, and hence stress loading of the megathrust earthquake zone, can occur in discrete pulses

https://science.sciencemag.org/content/sci/292/5521/1525.full.pdf

A silent slip event on the deeper Cascadia subduction interface.

The Multi-GNSS Experiment (MGEX) of the International GNSS Service (IGS) - Achievements, prospects and challenges

Bernese GPS Software Version 5.0

Bernese GNSS Software Version 5.2

Bernese GPS Software, Version 4.2

Since May 2003, the Center for Orbit Determination in Europe (CODE), one of the analysis centers of the International GNSS Service, has generated GPS and GLONASS products in a rigorous combined multi-system processing scheme, which promises the best possible consistency of the orbits of both systems. The resulting products, in particular the satellite orbits and clocks, are easily accessible by the user community. In the first part of this article, we focus on the generation of the combined global products at CODE, where we put emphasis not only on accuracy, but also on completeness. We study the impact of GLONASS on the CODE products, and the benefit of using them. Last, but not least, we introduce AGNES (Automated GNSS Network for Switzerland), a regional tracking network of small extensions (roughly 400 km East–West, 200 km North–South), which consequently tracks all GNSS satellites and analyzes their measurements using the CODE products.

/pdf/gnss-processing-at-code-status-report-1syp57hu5q.pdf

Rolf Dach

Papers

Bernese GPS Software Version 5.0

Bernese GNSS Software Version 5.2

Bernese GPS Software, Version 4.2

GNSS processing at CODE: status report

Technical Report 2012