Deepest Features Research Articles

Effective high-to-low-level feature aggregation network forendoscopic image classification.

The accuracy improvement in endoscopic image classification matters to the endoscopists in diagnosing and choosing suitable treatment for patients. Existing CNN-based methods for endoscopic image classification tend to use the deepest abstract features without considering the contribution of low-level features, while the latter is of great significance in the actual diagnosis of intestinal diseases. To make full use of both high-level and low-level features, we propose a novel two-stream network for endoscopic image classification. Specifically, the backbone stream is utilized to extract high-level features. In the fusion stream, low-level features are generated by a bottom-up multi-scale gradual integration (BMGI) method, and the input of BMGI is refined by top-down attention learning modules. Besides, a novel correction loss is proposed to clarify the relationship between high-level and low-level features. Experiments on the KVASIR dataset demonstrate that the proposed framework can obtain an overall classification accuracy of 97.33% with Kappa coefficient of 95.25%. Compared to the existing models, the two evaluation indicators have increased by 2% and 2.25%, respectively, at least. In this study, we proposed a two-stream network that fuses the high-level and low-level features for endoscopic image classification. The experiment results show that the high-to-low-level feature can better represent the endoscopic image and enable our model to outperform several state-of-the-art classification approaches. In addition, the proposed correction loss could regularize the consistency between backbone stream and fusion stream. Thus, the fused feature can reduce the intra-class distances and make accurate label prediction.

Salient object detection for RGB-D image by single stream recurrent convolution neural network

Salient object detection for RGB-D images aims to utilize color and depth information to automatically localize objects of human interest in the scene and reduce the complexity of visual analysis. Different from existing saliency detection model with double-stream network, salient object detection by Single Stream Recurrent Convolution Neural Network(SSRCNN) is proposed. First RGBD four-channels input is fed into VGG-16 net to generate multiple level features which express the most original feature for RGB-D image. The coarse saliency map from the deepest features can detect and localize salient objects, but loss the boundaries and subtle structures. So Depth Recurrent Convolution Neural Network (DRCNN) is then applied to each level feature for rendering salient object outline from deep to shallow hierarchically and progressively. With the help of deeper level feature, original depth cue and coarse saliency map, each level feature can accurately predict the salient objects in different scales. At last all the saliency maps from each level are fused together to generate final results. Extensive quantitative and qualitative experimental evaluations on four dataset demonstrate that the proposed method outperforms most state-of-the-art methods.

First glacier‐vibroseismic experiment ‐ results from cold firn of Colle Gnifetti

ABSTRACTIn the summer of 2010, a small shallow reflection seismic experiment was carried out on the firn covered cold glacier of Colle Gnifetti, Monte Rosa group, Swiss/Italian Alps. At this site, the physical properties of ice are comparable to polar conditions, which is why this site is often used for methodological tests. The experiment at 4500 m elevation was designed to explore the scope of shallow vibroseis for seismic targets within and below the glacier. A small ELVIS vibrator system was used to generate shear waves and compression waves for SH‐ and P‐wave receiver setups of two profiles. The resulting sections clearly show a boundary from ice to rock around 60 m and deeper structures below the glacier. The deepest features are estimated to be 150 m for the SH‐waves and 220 m for the P‐waves. Reflections could be detected also within the ice overburden, which are preliminarily interpreted as a change of density in the upper 30 m and possibly crystal orientation fabric in the ice column. Furthermore, elastic parameters could be derived from seismic velocities, due to clear basement reflections. The results of this unique experiment enable new insights into the internal structure of ice masses and open a promising new investigation method for sub‐ice structures and properties, such as basal sediments.

Implications of 21-cm observations for damped Ly systems

We present Giant Metrewave Radio Telescope HI 21cm absorption observations, of candidate and confirmed damped Lyman-$\alpha$ systems (DLAS). The derived spin temperatures (T_s) are in all cases $\sim 1000$ K or higher. We have also collated from the literature a list of DLAS for which 21cm observations exist, and discuss their implications for the nature of the absorbers. A cross-comparison of the 21cm profiles with low ionization metal profiles shows that the 21cm absorption coincides in velocity with the deepest metal line feature. This is consistent with models in which the deep metal line features arise from discrete clouds but not with models where the deepest features are the result of velocity crowding. We also find that the typical derived spin temperatures of DLAS are considerably higher than those in the Galaxy or nearby spirals. The only exceptions are DLAS which are known to be associated with the disks of spirals; these do, in fact, show low spin temperatures. In a multi-phase medium,the derived T_s is a weighted average of the temperatures of the individual phases. High derived T_s values are hence to be expected from small, low metallicity objects, since these objects should have a lower fraction of the cold phase in their ISM as compared to large galaxies. The high T_s in DLAS is hence consistent with their observed low metallicities as well as with recent observations that DLAS are also associated with dwarf/LSB galaxies. Finally, we suggest that the following trend may be identified: at low redshift, damped absorption arises from a range of systems, including spiral galaxy disks, while, at high redshift, absorption occurs predominantly in smaller systems. (Abridged)

Implications of 21-cm observations for damped Lyalpha systems

We present Giant Metrewave Radio Telescope H i 21-cm absorption observations, of candidate and confirmed damped Lyα systems. The derived spin temperatures Tspin are high, in all cases, ∼1000 K or higher. We have also collated from the published literature a list of damped absorbers for which 21-cm observations exist, and discuss the implications of the observations for the nature of these systems. A cross-comparison of the H i 21-cm profiles (which trace the cold gas) with the low ionization metal profiles (which trace all the neutral H i, both cold and warm) shows that in all cases for which both spectra are available, the 21-cm absorption coincides in velocity with the deepest metal line feature. This is consistent with models in which the deep metal line features arise from discrete clouds but not with models where the deepest features are the result of velocity crowding. We also find that the typical derived spin temperatures of damped Lyα systems are considerably higher than those in the Milky Way or nearby spiral galaxies. The only exceptions are systems that are known to be associated with the discs of spirals; these do, in fact, show low spin temperature. In a multi-phase medium the derived spin temperature is a weighted average of the temperatures of the individual phases. High apparent Tspin values are hence to be expected from small, low metallicity objects since these objects should (as per existing theoretical models of the formation of a multi-phase interstellar medium (ISM) in the Milky Way and high-redshift proto-galaxies) have a lower fraction of the cold phase in their ISM as compared with large galaxies. The high Tspin is hence consistent with the observed low metallicities of damped Lyα absorbers as well as with recent findings that damped absorption is associated with a variety of galaxy types (as opposed to being confined to the discs of large spirals). Finally, although the number of systems for which observations are available is small, we suggest that the following trend may be identified: at low redshift, damped Lyα absorption arises from a range of systems, including spiral galaxy discs, while, at high redshift, absorption occurs predominantly in smaller systems.

Time variations in the high-redshift 21 centimeter absorption spectrum of the BL Lacertae object AO 0235+164

view Abstract Citations (62) References (45) Co-Reads Similar Papers Volume Content Graphics Metrics Export Citation NASA/ADS Time variations in the high-redshift 21 CM absorption spectrum of theBL Lac object AO 235 +164. Wolfe, A. M. ; Davis, M. M. ; Briggs, F. H. Abstract We present evidence for time variations of the redshifted (Z =0.524)21 cm absorption spectrum of the BL Lac object AO 0235+164. A 4 year monitoring program with the 305 m telescope at Arecibo shows that the line depths of the four deepest features (A → D) vary in time. The strongest changes occur in feature D, which was observed to double its line depth in 0.5 yr. Weaker variations (between 15 and 35%) with durations between 0.25 and 0.5 yr are observed in all four features, and there is tentative evidence for a monthly flickering with amplitudes of ∼10%. Whereas each of the features evolves independently during some epochs, they appear to vary synchronously during others. We have also monitored the velocity centroids and velocity widths of the features and find that neither varies by more than ∼1 km s-1 km 5 1 We have placed a 3 σ upper limit of 0.75 km s-1 for centroid variations of feature D. But there is tentative evidence for systematic and correlated variations of ∼0.2 km s-1 in the velocity centroids of features A, B, and C. Throughout the monitoring program the continuum flux density at the observed frequency (932 MHz) of the 21 cm line has remained nearly constant. We interpret the line variations to stem from activity in AO 0235+164. The extrinsic model assumes that 0235, which has a redshift Z > 0.85, is at its redshift distance. When viewed through a nonuniform distribution of foreground clouds, variations in the brightness distribution of the background source cause apparent changes in line depth. This model predicts larger and more frequent shifts in velocity centroid than are observed. The intrinsic model assumes that 0235 is located near the absorbing gas, and that the redshift difference is attributed perhaps to an ejection velocity. Variations of the 21 cm continuum produce changes in spin temperature, and hence in 21 cm optical depth. While this model explains the synchronous changes in line depth and the absence of velocity variations rather naturally, the lack of a significant increase in 932 MHz flux during those epochs when the line depths decreased considerably is difficult to understand. We describe future tests which can rule out either of these models conclusively. Publication: The Astrophysical Journal Pub Date: August 1982 DOI: 10.1086/160187 Bibcode: 1982ApJ...259..495W Keywords: Absorption Spectra; Astronomical Models; Bl Lacertae Objects; Radio Sources (Astronomy); Red Shift; Centimeter Waves; Flux Density; Monte Carlo Method; Spectrum Analysis; Time Dependence; Astrophysics full text sources ADS | data products SIMBAD (1) NED (1)