SearcharxivSearch

arXiv subjects

Eric Hofesmann

Publications and source records attributed to Eric Hofesmann.

5 recordsLinked to original sources

ViP: Video Platform for PyTorch

This work presents the Video Platform for PyTorch (ViP), a deep learning-based framework designed to handle and extend to any problem domain based on videos. ViP supports (1) a single unified interface applicable to all video problem domains, (2) quick prototyping of video models, (3) executing large-batch operations with reduced memory consumption, and (4) easy and reproducible experimental setups. ViP's core functionality is built with flexibility and modularity in mind to allow for smooth data flow between different parts of the platform and benchmarking against existing methods. In providing a software platform that supports multiple video-based problem domains, we allow for more cross-pollination of models, ideas and stronger generalization in the video understanding research community.

cs.CV

Smoke and Mirrors: Signal-to-Noise and Time-Reversed Structures in Gamma-Ray Burst Pulse Light Curves

We demonstrate that the `smoke' of limited instrumental sensitivity smears out structure in gamma-ray burst (GRB) pulse light curves, giving each a triple-peaked appearance at moderate signal-to-noise and a simple monotonic appearance at low signal-to-noise. We minimize this effect by studying six very bright GRB pulses (signal-to-noise generally $> 100$), discovering surprisingly that each exhibits $\textit{complex time-reversible}$ wavelike residual structures. These `mirrored' wavelike structures can have large amplitudes, occur on short timescales, begin/end long before/after the onset of the monotonic pulse component, and have pulse spectra that generally evolve hard to soft, re-hardening at the time of each structural peak. Among other insights, these observations help explain the existence of negative pulse spectral lags, and allow us to conclude that GRB pulses are less common, more complex, and have longer durations than previously thought. Because structured emission mechanisms that can operate forwards and backwards in time seem unlikely, we look to $\textit{kinematic}$ behaviors to explain the time-reversed light curve structures. We conclude that each GRB pulse involves a single impactor interacting with an independent medium. Either the material is distributed in a bilaterally symmetric fashion, the impactor is structured in a bilaterally symmetric fashion, or the impactor's motion is reversed such that it returns along its original path of motion. The wavelike structure of the time-reversible component suggests that radiation is being both produced and absorbed/deflected dramatically, repeatedly, and abruptly from the monotonic component.

astro-ph.HE

M-PACT: An Open Source Platform for Repeatable Activity Classification Research

There are many hurdles that prevent the replication of existing work which hinders the development of new activity classification models. These hurdles include switching between multiple deep learning libraries and the development of boilerplate experimental pipelines. We present M-PACT to overcome existing issues by removing the need to develop boilerplate code which allows users to quickly prototype action classification models while leveraging existing state-of-the-art (SOTA) models available in the platform. M-PACT is the first to offer four SOTA activity classification models, I3D, C3D, ResNet50+LSTM, and TSN, under a single platform with reproducible competitive results. This platform allows for the generation of models and results over activity recognition datasets through the use of modular code, various preprocessing and neural network layers, and seamless data flow. In this paper, we present the system architecture, detail the functions of various modules, and describe the basic tools to develop a new model in M-PACT.

cs.CV

T-RECS: Training for Rate-Invariant Embeddings by Controlling Speed for Action Recognition

An action should remain identifiable when modifying its speed: consider the contrast between an expert chef and a novice chef each chopping an onion. Here, we expect the novice chef to have a relatively measured and slow approach to chopping when compared to the expert. In general, the speed at which actions are performed, whether slower or faster than average, should not dictate how they are recognized. We explore the erratic behavior caused by this phenomena on state-of-the-art deep network-based methods for action recognition in terms of maximum performance and stability in recognition accuracy across a range of input video speeds. By observing the trends in these metrics and summarizing them based on expected temporal behaviour w.r.t. variations in input video speeds, we find two distinct types of network architectures. In this paper, we propose a preprocessing method named T-RECS, as a way to extend deep-network-based methods for action recognition to explicitly account for speed variability in the data. We do so by adaptively resampling the inputs to a given model. T-RECS is agnostic to the specific deep-network model; we apply it to four state-of-the-art action recognition architectures, C3D, I3D, TSN, and ConvNet+LSTM. On HMDB51 and UCF101, T-RECS-based I3D models show a peak improvement of at least 2.9% in performance over the baseline while T-RECS-based C3D models achieve a maximum improvement in stability by 59% over the baseline, on the HMDB51 dataset.

cs.CV

Properties of Short Gamma-Ray Burst Pulses from A BATSE TTE GRB Pulse Catalog

We analyze pulse properties of Short gamma-ray bursts (GRBs) from a new catalog containing 434 pulses from 387 BATSE Time-Tagged Event (TTE) GRBs. Short GRB pulses exhibit correlated properties of duration, fluence, hardness, and amplitude, and they evolve hard-to-soft while undergoing similar triple-peaked light curves similar to those found in Long/Intermediate bursts. We classify pulse light curves using their temporal complexities, demonstrating that Short GRB pulses exhibit a range of complexities from smooth to highly variable. Most of the bright, hard, chaotic emission seen in complex pulses seems to represent a separate highly-variable emission component. Unlike Long/Intermediate bursts, as many as 90\% of Short GRBs are single-pulsed. However, emission in Short multi-pulsed bursts is coupled such that the first pulse's duration is a predictor of both the interpulse separation and subsequent pulse durations. These results strongly support the idea that external shocks produce the prompt emission seen in Short GRBs. The similarities between the triple-peaked structures and spectral evolution of Long, Short, and Intermediate GRBs then suggests that external shocks are responsible for the prompt emission observed in all GRB classes. In addition to these findings, we identify a new type of gamma-ray transient in which peak amplitudes occur at the end of the burst rather than at earlier times. Some of these "Crescendo" bursts are preceded by rapid-fire "Staccato" pulses, whereas the remaining are preceded by a variable episode that could be unresolved staccato pulses.

astro-ph.HE