88 research outputs found
Anticipating Next Active Objects for Egocentric Videos
Get PDFThis paper addresses the problem of anticipating the next-active-object
location in the future, for a given egocentric video clip where the contact
might happen, before any action takes place. The problem is considerably hard,
as we aim at estimating the position of such objects in a scenario where the
observed clip and the action segment are separated by the so-called ``time to
contact'' (TTC) segment. Many methods have been proposed to anticipate the
action of a person based on previous hand movements and interactions with the
surroundings. However, there have been no attempts to investigate the next
possible interactable object, and its future location with respect to the
first-person's motion and the field-of-view drift during the TTC window. We
define this as the task of Anticipating the Next ACTive Object (ANACTO). To
this end, we propose a transformer-based self-attention framework to identify
and locate the next-active-object in an egocentric clip.
We benchmark our method on three datasets: EpicKitchens-100, EGTEA+ and
Ego4D. We also provide annotations for the first two datasets. Our approach
performs best compared to relevant baseline methods. We also conduct ablation
studies to understand the effectiveness of the proposed and baseline methods on
varying conditions. Code and ANACTO task annotations will be made available
upon paper acceptance.Comment: Accepted by IEEE ACCESS, this paper carries the Manuscript DOI:
10.1109/ACCESS.2024.3395282. The complete peer-reviewed version is available
via this DOI, while the arXiv version is a post-author manuscript without
peer-revie
Exploring the rainfall data from satellites to monitor rainfall induced landslides – A case study
No full textIn the present study, rainfall estimates from TRMM (Tropical Rainfall Measuring Mission) and GPM (Global Precipitation Mission) constellation of satellites are analyzed in the context of rainfall induced landslide occurrences over Western Ghats (WG) of India along with the daily gridded rainfall data developed by the India Meteorological Department (IMD) and Advanced Research Weather Research and Forecasting (ARW) numerical model simulations. This study aims to analyze the pattern of changes in rain rate and total rainfall triggering the large landslides over WG in TMPA (product of TRMM) and IMERG (GPM product) rainfall data sets. As a case study, performance of IMERG V5 is assessed during Malin landslide which occurred on 30 July 2014 (initial GPM era). Results indicate that IMERG shows significant increase in rain rate (>60 mm/h in half-hourly data) during Malin landslide. Near real-time IMERG V5, underestimates the rain rate but increasing pattern of rain-rate are observed which is similar to that of final version. Spatial pattern of ARW rainfall output is also close to the satellite and IMD rainfall patterns. We propose that IMERG V5 can be used as an indicator to reliably depict the higher rainfall scenario over the sites that are vulnerable to rainfall induced landslide occurrence over the WG region.Authors are thankful to the anonymous reviewers for their constructive comments that improved the quality of the article. The author Mr. Manoj Kumar Thakur is thankful to Silver Jubilee Scholarship Scheme-Govt. of India under which he is sponsored
Back calculation of Debris flow Run-Out & Entrainment Using the Voellmy Rheology
Get PDFDebris flow is one of the many geo-hazards that cause a major damage worldwide. It can cause loss of human lives especially to those living in mountainous regions. Besides, it cause economic damage by destroying properties and infrastructure. Forecasting and controlling the hazard associated to this type of mass movements is still a difficult task that requires qualitative and quantitative analyses. However, the development of numerical dynamic run out models has a major advantage in the study of this processes, as they allow the simulation of possible future scenarios. Some of these numerical models currently in use for simulating debris flows are MassMov2D, DAN-3D, FLO-2D and RAMMS.
The main objective of this thesis was to back calculate debris flow run out and its entrainment behavior using numerical models. For this study, the Author uses RAMMS run out model for the back analysis of debris flow mobility. RAMMS is able to model entrainment along the flow path by using rate controlled entrainment method, which regulates the amount of mass being entrained in to the debris flow and the time needed to accelerate this mass to the debris flow velocity.
One Norwegian debris flow, Mjåland debris flow happened in June 2016, was back calculated using RAMMS. The Voellmy rheological model was used first to calibrate the input parameters (friction coefficient, turbulence factor and entrainment coefficient) and then to test the sensitivity of each parameter. The model is found to be highly sensitive to entrainment coefficient, K, friction coefficient, µ and turbulent friction, ξ.
The result of this study also showed that the velocity and height of the flow with entrainment and slope is relatively greater than the normal range for debris flow. Although RAMMS was able to simulate entrainment, the quality of its output depends on the resolution of digital elevation model (DEM) used as input during modelling
Back calculation of Debris flow Run-Out & Entrainment Using the Voellmy Rheology
No full textDebris flow is one of the many geo-hazards that cause a major damage worldwide. It can cause loss of human lives especially to those living in mountainous regions. Besides, it cause economic damage by destroying properties and infrastructure. Forecasting and controlling the hazard associated to this type of mass movements is still a difficult task that requires qualitative and quantitative analyses. However, the development of numerical dynamic run out models has a major advantage in the study of this processes, as they allow the simulation of possible future scenarios. Some of these numerical models currently in use for simulating debris flows are MassMov2D, DAN-3D, FLO-2D and RAMMS.
The main objective of this thesis was to back calculate debris flow run out and its entrainment behavior using numerical models. For this study, the Author uses RAMMS run out model for the back analysis of debris flow mobility. RAMMS is able to model entrainment along the flow path by using rate controlled entrainment method, which regulates the amount of mass being entrained in to the debris flow and the time needed to accelerate this mass to the debris flow velocity.
One Norwegian debris flow, Mjåland debris flow happened in June 2016, was back calculated using RAMMS. The Voellmy rheological model was used first to calibrate the input parameters (friction coefficient, turbulence factor and entrainment coefficient) and then to test the sensitivity of each parameter. The model is found to be highly sensitive to entrainment coefficient, K, friction coefficient, µ and turbulent friction, ξ.
The result of this study also showed that the velocity and height of the flow with entrainment and slope is relatively greater than the normal range for debris flow. Although RAMMS was able to simulate entrainment, the quality of its output depends on the resolution of digital elevation model (DEM) used as input during modelling
Analyses of contact networks of community dogs on a university campus in Nakhon Pathom, Thailand
No full textFree-roaming dogs have been identified as an important reservoir of rabies in many countries including Thailand. There is a need for novel insights to improve current rabies control strategies in these countries. Network analysis is commonly used to study the interactions between individuals or organizations and has been applied in preventive veterinary medicine. However, contact networks of domestic free-roaming dogs are mostly unexplored. The objective of this study was to explore the contact network of free-roaming dogs residing on a university campus. Three one-mode networks were created using co-appearances of dogs as edges. A two-mode network was created by associating the dog with the pre-defined area it was seen in. The average number of contacts a dog had was 6.74. The normalized degree for the weekend network was significantly higher compared to the weekday network. All one-mode networks displayed small-world network characteristics. Most dogs were observed in only one area. The average number of dogs which shared an area was 8.67. In this study, we demonstrated the potential of observational methods to create networks of contacts. The network information acquired can be further used in network modeling and designing targeted disease control programs
- …
