An Efficient Multi-Scale Anchor Box Approach to Detect Partial Faces from a Video Sequence

In recent years, face detection has achieved considerable attention in the field of computer vision using traditional machine learning techniques and deep learning techniques. Deep learning is used to build the most recent and powerful face detection algorithms. However, partial face detection still remains to achieve remarkable performance. Partial faces are occluded due to hair, hat, glasses, hands, mobile phones, and side-angle-captured images. Fewer facial features can be identified from such images. In this paper, we present a deep convolutional neural network face detection method using the anchor boxes section strategy. We limited the number of anchor boxes and scales and chose only relevant to the face shape. The proposed model was trained and tested on a popular and challenging face detection benchmark dataset, i.e., Face Detection Dataset and Benchmark (FDDB), and can also detect partially covered faces with better accuracy and precision. Extensive experiments were performed, with evaluation metrics including accuracy, precision, recall, F1 score, inference time, and FPS. The results show that the proposed model is able to detect the face in the image, including occluded features, more precisely than other state-of-the-art approaches, achieving 94.8% accuracy and 98.7% precision on the FDDB dataset at 21 frames per second (FPS).

An Efficient Deep Convolutional Neural Network Approach for Object Detection and Recognition Using a Multi-Scale Anchor Box in Real-Time

Deep learning is a relatively new branch of machine learning in which computers are taught to recognize patterns in massive volumes of data. It primarily describes learning at various levels of representation, which aids in understanding data that includes text, voice, and visuals. Convolutional neural networks have been used to solve challenges in computer vision, including object identification, image classification, semantic segmentation and a lot more. Object detection in videos involves confirming the presence of the object in the image or video and then locating it accurately for recognition. In the video, modelling techniques suffer from high computation and memory costs, which may decrease performance measures such as accuracy and efficiency to identify the object accurately in real-time. The current object detection technique based on a deep convolution neural network requires executing multilevel convolution and pooling operations on the entire image to extract deep semantic properties from it. For large objects, detection models can provide superior results; however, those models fail to detect the varying size of the objects that have low resolution and are greatly influenced by noise because the features after the repeated convolution operations of existing models do not fully represent the essential characteristics of the objects in real-time. With the help of a multi-scale anchor box, the proposed approach reported in this paper enhances the detection accuracy by extracting features at multiple convolution levels of the object. The major contribution of this paper is to design a model to understand better the parameters and the hyper-parameters which affect the detection and the recognition of objects of varying sizes and shapes, and to achieve real-time object detection and recognition speeds by improving accuracy. The proposed model has achieved 84.49 mAP on the test set of the Pascal VOC-2007 dataset at 11 FPS, which is comparatively better than other real-time object detection models.

Metrological traceability of measurement results and calibration hierarchy is a prerequisite for improved data compatibility: example of the VPDB scale.

<p>For stable isotope data sets to be compared or combined in biogeochemical studies, their compatibility must be well understood. For δ13C measurements in greenhouse gases, the WMO GAW program has set compatibility targets of 0.010 ‰ for atmospheric CO2 and 0.020 ‰ for atmospheric methane (in background air studies [1, 2]). The direct comparison of samples between laboratories can provide limited information, such as a snapshot for a specific time period, but combining data sets produced over decades requires more efforts. To produce high quality data, reliable calibrations must be made, mutually consistent values of reference materials (RMs) must be used, and a traceability scheme that ensures low uncertainty must be implemented.</p><p>The VPDB δ13C scale provides example of approaches developed recently. Several problems with the existing implementation of the VPDB scale have been identified between 2009-2016 [3]: the primary reference material (RM) NBS19 was exhausted and needed to be replaced; the δ13C of LSVEC (used to anchor the VPDB scale at negative δ13C) was found to be drifting and its use as a RM for δ13C was discontinued [4]; other RMs that were available in 2016 (e.g., NBS18) were not able to be used to develop new RMs as their uncertainties were too large. Given that the VPDB scale is artefact-based and not supported by absolute ratio measurements with uncertainty as low as required, the principles of value assignments on the VPDB scale were needed to be revised.</p><p>To ensure that a revised scheme did not encounter similar problems (with dependence on a single scale-anchor), several fundamental metrological principles were considered: (i) traceability of measurement results to the primary RM, (ii) a hierarchy of calibrators and (iii) comprehensive understanding of measurement method(s) [5]. The revised VPDB scheme [3] was applied to the new primary RM [6] and three RMs covering a large δ13C range (to negative values) [7]. Values were assigned in a mutually consistent way, with uncertainties ranging from 0.010 to 0.015 ‰, depending on the assigned δ13C. Each RM value has an uncertainty assigned that includes all known instrumental corrections, potential alterations due to storage, and inhomogeneity assessment [6,7]. The scheme allows for the δ13C range to be expanded by developing new carbonate RMs, and to be extended to matrix-based RMs.</p><p>The revised VPDB δ13C scale realization should lead to a robust basis for improving data compatibility. The developed framework can be applied to other measurements of biogeochemical interest, such as small 17O variations (in H2O, carbonates and other samples), clumped isotopes, and various paleoclimate reconstructions. Notably, the traceability principle is helpful in realistic uncertainty estimations which provide a tool to understand constrains and limiting steps in data comparisons.</p><p>REFERENCES:  [1]. WMO, GAW Report No.229. 2016. [2]. WMO, GAW Report No.242. 2018. [3]. Assonov, S. et al., RCM, 2021. https://doi.org/10.1002/rcm.9018. [4]. IUPAC, Press release of the IUPAC meeting in 2017, https://iupac.org/standard-atomic-weights-of-14-chemical-elements-revised/. [5]. De Bievre, P. et al., PURE APPL CHEM, 2011. <strong>83</strong>(10): p. 1873-1935. [6]. Assonov, S., et al., RCM, 2020: p. https://doi.org/10.1002/rcm.8867. [7]. Assonov, S. et al., RCM, 2021. https://doi.org/10.1002/rcm.9014</p>

The IAEA carbonate reference materials aimed at the VPDB scale realization with low uncertainty.

<p>The stable isotope scales of the light elements (H, C, O, S) are artefact-based (related to a primary reference material) and their practical realisation is based on several refence materials (RMs) traceable to the primary RM on a respective delta-scale. NBS19 carbonate, the primary RM for the VPDB scale introduced in 1987, exhausted in 2012, and its replacement was not available for several years. In 2016, IAEA-603 carbonate (replacement for NBS19) was released as the new primary RM having been carefully calibrated versus the remaining NBS19. The IAEA-603 uncertainty in <span>δ</span>13C and <span>δ</span>18O for the first batch (5200 ampoules produced) is ±0.010 ‰ and ±0.040 ‰ respectively (1-sigma level); the homogeneity assessment is the major component of total uncertainty which is limited by the best mass-spectrometer performance and the method (carbonate-acid reaction) reproducibility.</p><p>In 2015, monitoring of LSVEC (formerly the second scale-anchor on the VPDB scale) detected variable drifts in its <span>δ</span>13C value and therefore the use of LSVEC as RM for <span>δ</span>13C was discontinued. It was recognised that a replacement for LSVEC is needed for normalization of the <span>δ</span>13C measurement results, also to address the strict uncertainty requirements for <span>δ</span>13C observations in atmospheric CO2 and methane (≤0.01 ‰ and ≤0.02 ‰ correspondingly). Similar to IAEA-603, any new RMs will address the technical requirements for RMs laid out by ISO Guide 35: 2017 including (i) RM batch production and batch characterisation; (ii) homogeneity and stability assessment of the final product (RMs sealed off in 0.5 g ampoules) and (iii) value and uncertainty assignment based on the metrological traceability. Three new carbonate RMs are in preparation at the IAEA; the uncertainty in <span>δ</span>13C for all three materials due to RM’ homogeneity is already confirmed at ≤0.01 ‰ (on 10 mg aliquots), which is at the limit of the best modern mass-spectrometers. The isotopic characterisation of these new carbonate RMs is in progress; they should be released in 2020.</p><p>Together with IAEA-603, the three new RMs will provide a reliable realization of the VPDB scale with the lowest possible uncertainty. With these RMs users can (i) select RMs in a suitable <span>δ</span>13C range, (ii) detect any potential drift of RMs including the behaviour of daily lab-standards and (iii) detect any potential problem in applying the 17O correction at end-user laboratories. In conclusion, these new reference materials will allow laboratories worldwide to establish metrological comparability for decades.</p>

Development and evaluation of a suite of isotope reference gases for methane in air

Measurements from multiple laboratories have to be related to unifying and traceable reference material in order to be comparable. However, such fundamental reference materials are not available for isotope ratios in atmospheric methane, which led to misinterpretations of combined data sets in the past. We developed a method to produce a suite of synthetic CH4-in-air standard gases that can be used to unify methane isotope ratio measurements of laboratories in the atmospheric monitoring community. Therefore, we calibrated a suite of pure methane gases of different methanogenic origin against international referencing materials that define the VSMOW (Vienna Standard Mean Ocean Water) and VPDB (Vienna Pee Dee Belemnite) isotope scales. The isotope ratios of our pure methane gases range between −320 and +40 ‰ for δ2H–CH4 and between −70 and −40 ‰ for δ13C–CH4, enveloping the isotope ratios of tropospheric methane (about −85 and −47 ‰ for δ2H–CH4 and δ13C–CH4 respectively). Estimated uncertainties, including the full traceability chain, are < 1.5 ‰ and < 0.2 ‰ for δ2H and δ13C calibrations respectively. Aliquots of the calibrated pure methane gases have been diluted with methane-free air to atmospheric methane levels and filled into 5 L glass flasks. The synthetic CH4-in-air standards comprise atmospheric oxygen/nitrogen ratios as well as argon, krypton and nitrous oxide mole fractions to prevent gas-specific measurement artefacts. The resulting synthetic CH4-in-air standards are referred to as JRAS-M16 (Jena Reference Air Set – Methane 2016) and will be available to the atmospheric monitoring community. JRAS-M16 may be used as unifying isotope scale anchor for isotope ratio measurements in atmospheric methane, so that data sets can be merged into a consistent global data frame.

Jena Reference Air Set (JRAS): a multi-point scale anchor for isotope measurements of CO₂ in air

The need for a unifying scale anchor for isotopes of CO2 in air was brought to light at the 11th WMO/IAEA Meeting of Experts on Carbon Dioxide in Tokyo 2001. During discussions about persistent discrepancies in isotope measurements between the worlds leading laboratories, it was concluded that a unifying scale anchor for Vienna Pee Dee Belemnite (VPDB) of CO2 in air was desperately needed. Ten years later, at the 2011 Meeting of Experts on Carbon Dioxide in Wellington, it was recommended that the Jena Reference Air Set (JRAS) become the official scale anchor for isotope measurements of CO2 in air (Brailsford, 2012). The source of CO2 used for JRAS is two calcites. After releasing CO2 by reaction with phosphoric acid, the gases are mixed into CO2-free air. This procedure ensures both isotopic stability and longevity of the CO2. That the reference CO2 is generated from calcites and supplied as an air mixture is unique to JRAS. This is made to ensure that any measurement bias arising from the extraction procedure is eliminated. As every laboratory has its own procedure for extracting the CO2, this is of paramount importance if the local scales are to be unified with a common anchor. For a period of four years, JRAS has been evaluated through the IMECC1 program, which made it possible to distribute sets of JRAS gases to 13 laboratories worldwide. A summary of data from the six laboratories that have reported the full set of results is given here along with a description of the production and maintenance of the JRAS scale anchors. 1 IMECC refers to the EU project "Infrastructure for Measurements of the European Carbon Cycle" (http://imecc.ipsl.jussieu.fr/).

Jena Reference Air Set (JRAS): a multi-point scale anchor for isotope measurements of CO₂ in air

The need for a unifying scale anchor for isotopes of CO2 in air was brought to light at the WMO CO2 Experts Meeting in Tokyo 2001. During discussions about persistent discrepancies in isotope measurements between the worlds leading laboratories it was concluded that a unifying scale anchor for VPDB of CO2 in air was desperately needed. Now, 10 yr later, the 2011 CO2-Experts-Meeting in Wellington has decided that the Jena Reference Air Set or JRAS is recommended as official scale anchor for isotope measurements of CO2 in air. The JRAS gases consist of reference CO2 mixed into CO2 free air. To safeguard both stability and longevity of the CO2, it is directly generated from two solid reference calcites. That the reference CO2 is supplied in air is unique to JRAS. This is made to ensure that any measurement bias arising from the extraction procedure is eliminated. As every lab has its own procedure for extracting the CO2 this is of paramount importance if the local scales are to be unified. For a period of four years, JRAS has been evaluated through the IMECC program, which made it possible to distribute sets of JRAS gases to 11 laboratories worldwide. A summary of the results is reported here along with a description of the production and maintenance of the JRAS scale anchors.

Analysis of anchor foundation with root caissons loaded in nonhomogeneous soils

This paper presents a novel type of deep foundation — an anchor foundation with root caissons, which improve the soil–caisson interaction using penetrating roots. A simplified analytical procedure is proposed for calculating responses of the new anchor foundation subjected to combined vertical, horizontal, and moment loadings. Caisson–soil interaction is simulated by a Winkler subgrade reaction model, with an improved Winkler spring constant to consider the reinforcement effect of roots. The method incorporates limiting soil stress to investigate soil nonlinear behavior and a transfer matrix to represent nonhomogeneous soils, with a passive caisson model accounting for group effects. The effectiveness of the present approach is evaluated by comparison with full-scale field load tests and a finite element analysis on a small-scale anchor foundation to give a preliminary design guideline for the Ma’anshan Yangtze River Bridge in China.