Showing papers on "Detector published in 2022"

Detecting Twenty-Thousand Classes Using Image-Level Supervision

Abstract: Current object detectors are limited in vocabulary size due to the small scale of detection datasets. Image classifiers, on the other hand, reason about much larger vocabularies, as their datasets are larger and easier to collect. We propose Detic, which simply trains the classifiers of a detector on image classification data and thus expands the vocabulary of detectors to tens of thousands of concepts. Unlike prior work, Detic does not need complex assignment schemes to assign image labels to boxes based on model predictions, making it much easier to implement and compatible with a range of detection architectures and backbones. Our results show that Detic yields excellent detectors even for classes without box annotations. It outperforms prior work on both open-vocabulary and long-tail detection benchmarks. Detic provides a gain of 2.4 mAP for all classes and 8.3 mAP for novel classes on the open-vocabulary LVIS benchmark. On the standard LVIS benchmark, Detic obtains 41.7 mAP when evaluated on all classes, or only rare classes, hence closing the gap in performance for object categories with few samples. For the first time, we train a detector with all the twenty-one-thousand classes of the ImageNet dataset and show that it generalizes to new datasets without finetuning. Code is available at https://github.com/facebookresearch/Detic .

First Clinical Photon-counting Detector CT System: Technical Evaluation

Abstract: Background The first clinical CT system to use photon-counting detector (PCD) technology has become available for patient care. Purpose To assess the technical performance of the PCD CT system with use of phantoms and representative participant examinations. Materials and Methods Institutional review board approval and written informed consent from four participants were obtained. Technical performance of a dual-source PCD CT system was measured for standard and high-spatial-resolution (HR) collimations. Noise power spectrum, modulation transfer function, section sensitivity profile, iodine CT number accuracy in virtual monoenergetic images (VMIs), and iodine concentration accuracy were measured. Four participants were enrolled (between May 2021 and August 2021) in this prospective study and scanned using similar or lower radiation doses as their respective clinical examinations performed on the same day using energy-integrating detector (EID) CT. Image quality and findings from the participants' PCD CT and EID CT examinations were compared. Results All standard technical performance measures met accreditation and regulatory requirements. Relative to filtered back-projection reconstructions, images from iterative reconstruction had lower noise magnitude but preserved noise power spectrum shape and peak frequency. Maximum in-plane spatial resolutions of 125 and 208 µm were measured for HR and standard PCD CT scans, respectively. Minimum values for section sensitivity profile full width at half maximum measurements were 0.34 mm (0.2-mm nominal section thickness) and 0.64 mm (0.4-mm nominal section thickness) for HR and standard PCD CT scans, respectively. In a 120-kV standard PCD CT scan of a 40-cm phantom, VMI iodine CT numbers had a mean percentage error of 5.7%, and iodine concentration had root mean squared error of 0.5 mg/cm3, similar to previously reported values for EID CT. VMIs, iodine maps, and virtual noncontrast images were created for a coronary CT angiogram acquired with 66-msec temporal resolution. Participant PCD CT images showed up to 47% lower noise and/or improved spatial resolution compared with EID CT. Conclusion Technical performance of clinical photon-counting detector (PCD) CT is improved relative to that of a current state-of-the-art CT system. The dual-source PCD geometry facilitated 66-msec temporal resolution multienergy cardiac imaging. Study participant images illustrated the effect of the improved technical performance. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Willemink and Grist in this issue.

Anchor DETR: Query Design for Transformer-Based Detector

Abstract: In this paper, we propose a novel query design for the transformer-based object detection. In previous transformer-based detectors, the object queries are a set of learned embeddings. However, each learned embedding does not have an explicit physical meaning and we cannot explain where it will focus on. It is difficult to optimize as the prediction slot of each object query does not have a specific mode. In other words, each object query will not focus on a specific region. To solve these problems, in our query design, object queries are based on anchor points, which are widely used in CNN-based detectors. So each object query focuses on the objects near the anchor point. Moreover, our query design can predict multiple objects at one position to solve the difficulty: ``one region, multiple objects''. In addition, we design an attention variant, which can reduce the memory cost while achieving similar or better performance than the standard attention in DETR. Thanks to the query design and the attention variant, the proposed detector that we called Anchor DETR, can achieve better performance and run faster than the DETR with 10x fewer training epochs. For example, it achieves 44.2 AP with 19 FPS on the MSCOCO dataset when using the ResNet50-DC5 feature for training 50 epochs. Extensive experiments on the MSCOCO benchmark prove the effectiveness of the proposed methods. Code is available at https://github.com/megvii-research/AnchorDETR.

Comparing scientific abstracts generated by ChatGPT to original abstracts using an artificial intelligence output detector, plagiarism detector, and blinded human reviewers

Abstract: Background Large language models such as ChatGPT can produce increasingly realistic text, with unknown information on the accuracy and integrity of using these models in scientific writing. Methods We gathered ten research abstracts from five high impact factor medical journals (n=50) and asked ChatGPT to generate research abstracts based on their titles and journals. We evaluated the abstracts using an artificial intelligence (AI) output detector, plagiarism detector, and had blinded human reviewers try to distinguish whether abstracts were original or generated. Results All ChatGPT-generated abstracts were written clearly but only 8% correctly followed the specific journal’s formatting requirements. Most generated abstracts were detected using the AI output detector, with scores (higher meaning more likely to be generated) of median [interquartile range] of 99.98% [12.73, 99.98] compared with very low probability of AI-generated output in the original abstracts of 0.02% [0.02, 0.09]. The AUROC of the AI output detector was 0.94. Generated abstracts scored very high on originality using the plagiarism detector (100% [100, 100] originality). Generated abstracts had a similar patient cohort size as original abstracts, though the exact numbers were fabricated. When given a mixture of original and general abstracts, blinded human reviewers correctly identified 68% of generated abstracts as being generated by ChatGPT, but incorrectly identified 14% of original abstracts as being generated. Reviewers indicated that it was surprisingly difficult to differentiate between the two, but that the generated abstracts were vaguer and had a formulaic feel to the writing. Conclusion ChatGPT writes believable scientific abstracts, though with completely generated data. These are original without any plagiarism detected but are often identifiable using an AI output detector and skeptical human reviewers. Abstract evaluation for journals and medical conferences must adapt policy and practice to maintain rigorous scientific standards; we suggest inclusion of AI output detectors in the editorial process and clear disclosure if these technologies are used. The boundaries of ethical and acceptable use of large language models to help scientific writing remain to be determined.

Lite-YOLOv5: A Lightweight Deep Learning Detector for On-Board Ship Detection in Large-Scene Sentinel-1 SAR Images

Abstract: Synthetic aperture radar (SAR) satellites can provide microwave remote sensing images without weather and light constraints, so they are widely applied in the maritime monitoring field. Current SAR ship detection methods based on deep learning (DL) are difficult to deploy on satellites, because these methods usually have complex models and huge calculations. To solve this problem, based on the You Only Look Once version 5 (YOLOv5) algorithm, we propose a lightweight on-board SAR ship detector called Lite-YOLOv5, which (1) reduces the model volume; (2) decreases the floating-point operations (FLOPs); and (3) realizes the on-board ship detection without sacrificing accuracy. First, in order to obtain a lightweight network, we design a lightweight cross stage partial (L-CSP) module to reduce the amount of calculation and we apply network pruning for a more compact detector. Then, in order to ensure the excellent detection performance, we integrate a histogram-based pure backgrounds classification (HPBC) module, a shape distance clustering (SDC) module, a channel and spatial attention (CSA) module, and a hybrid spatial pyramid pooling (H-SPP) module to improve detection performance. To evaluate the on-board SAR ship detection ability of Lite-YOLOv5, we also transplant it to the embedded platform NVIDIA Jetson TX2. Experimental results on the Large-Scale SAR Ship Detection Dataset-v1.0 (LS-SSDD-v1.0) show that Lite-YOLOv5 can realize lightweight architecture with a 2.38 M model volume (14.18% of model size of YOLOv5), on-board ship detection with a low computation cost (26.59% of FLOPs of YOLOv5), and superior detection accuracy (1.51% F1 improvement compared with YOLOv5).

Synergistic strain engineering of perovskite single crystals for highly stable and sensitive X-ray detectors with low-bias imaging and monitoring

Abstract: Abstract Although three-dimensional metal halide perovskite (ABX 3 ) single crystals are promising next-generation materials for radiation detection, state-of-the-art perovskite X-ray detectors include methylammonium as A-site cations, limiting the operational stability. Previous efforts to improve the stability using formamidinium–caesium-alloyed A-site cations usually sacrifice the detection performance because of high trap densities. Here we successfully solve this trade-off between stability and detection performance by synergistic composition engineering, where we include A-site alloys to decrease the trap density and B-site dopants to release the microstrain induced by A-site alloying. As such, we develop high-performance perovskite X-ray detectors with excellent stability. Our X-ray detectors exhibit high sensitivity of (2.6 ± 0.1) × 10 4 μC Gy air −1 cm −2 under 1 V cm −1 and ultralow limit of detection of 7.09 nGy air s −1 . In addition, they feature long-term operational stability over half a year and impressive thermal stability up to 125 °C. We further demonstrate the promise of our perovskite X-ray detectors for low-bias portable applications with high-quality X-ray imaging and monitoring prototypes.

Object detection using YOLO: challenges, architectural successors, datasets and applications

Abstract: Object detection is one of the predominant and challenging problems in computer vision. Over the decade, with the expeditious evolution of deep learning, researchers have extensively experimented and contributed in the performance enhancement of object detection and related tasks such as object classification, localization, and segmentation using underlying deep models. Broadly, object detectors are classified into two categories viz. two stage and single stage object detectors. Two stage detectors mainly focus on selective region proposals strategy via complex architecture; however, single stage detectors focus on all the spatial region proposals for the possible detection of objects via relatively simpler architecture in one shot. Performance of any object detector is evaluated through detection accuracy and inference time. Generally, the detection accuracy of two stage detectors outperforms single stage object detectors. However, the inference time of single stage detectors is better compared to its counterparts. Moreover, with the advent of YOLO (You Only Look Once) and its architectural successors, the detection accuracy is improving significantly and sometime it is better than two stage detectors. YOLOs are adopted in various applications majorly due to their faster inferences rather than considering detection accuracy. As an example, detection accuracies are 63.4 and 70 for YOLO and Fast-RCNN respectively, however, inference time is around 300 times faster in case of YOLO. In this paper, we present a comprehensive review of single stage object detectors specially YOLOs, regression formulation, their architecture advancements, and performance statistics. Moreover, we summarize the comparative illustration between two stage and single stage object detectors, among different versions of YOLOs, applications based on two stage detectors, and different versions of YOLOs along with the future research directions.

Nonlinear photoresponse of high damage threshold titanium disulfide nanocrystals for Q-switched pulse generation

Abstract: A passively Q-switched Er-doped fiber laser with two wavelength at 1530 nm and 1556 nm wavelength was designed based on the titanium disulfide (TiS2) saturable absorber, the ultra-fast modulator was fabricated with TiS2 by the liquid-phase exfoliation and spin-coating methods and transferred onto a fiber ferrule. In the experiment, two single-mode 980 nm pump sources were applied in order to detect the high damage threshold of TiS2. The shortest pulse duration and maximum output power of the stable self-starting Q-switched fiber laser are 1.45 µs and 3.93 mW, respectively. The adjustable range of the repetition rate is from 25.8 kHz to 126.8 kHz with about 100 kHz tuning range, and the pulse with energy up to 39.3 nJ. Our experimental results conclusively suggest that TiS2 nanocrystals were advanced nanomaterial with high damage threshold would have extensive application prospects in the field of pulse fiber lasers.

MSFT-YOLO: Improved YOLOv5 Based on Transformer for Detecting Defects of Steel Surface

Abstract: With the development of artificial intelligence technology and the popularity of intelligent production projects, intelligent inspection systems have gradually become a hot topic in the industrial field. As a fundamental problem in the field of computer vision, how to achieve object detection in the industry while taking into account the accuracy and real-time detection is an important challenge in the development of intelligent detection systems. The detection of defects on steel surfaces is an important application of object detection in the industry. Correct and fast detection of surface defects can greatly improve productivity and product quality. To this end, this paper introduces the MSFT-YOLO model, which is improved based on the one-stage detector. The MSFT-YOLO model is proposed for the industrial scenario in which the image background interference is great, the defect category is easily confused, the defect scale changes a great deal, and the detection results of small defects are poor. By adding the TRANS module, which is designed based on Transformer, to the backbone and detection headers, the features can be combined with global information. The fusion of features at different scales by combining multi-scale feature fusion structures enhances the dynamic adjustment of the detector to objects at different scales. To further improve the performance of MSFT-YOLO, we also introduce plenty of effective strategies, such as data augmentation and multi-step training methods. The test results on the NEU-DET dataset show that MSPF-YOLO can achieve real-time detection, and the average detection accuracy of MSFT-YOLO is 75.2, improving about 7% compared to the baseline model (YOLOv5) and 18% compared to Faster R-CNN, which is advantageous and inspiring.

ArcNet: Series AC Arc Fault Detection Based on Raw Current and Convolutional Neural Network

Abstract: AC series arc is dangerous and can cause serious electric fire hazards and property damage. This article proposed a convolutional neural network -based arc detection model named ArcNet. The database of this research is collected from eight different types of loads according to IEC62606 standard. The two most common types of arcs, including arcs from a loose connection of cables and those caused by the failure of the insulation, are generated in testing and included in the database. Using the database of raw current, experimental results indicate ArcNet can achieve a maximum of 99.47% arc detection accuracy at 10 kHz sampling rate. The model is also implemented in Raspberry Pi 3B for classification accuracy. A tradeoff study between the arc detection accuracy and model runtime has been conducted. The proposed ArcNet obtained an average runtime of 31 ms/sample of 1 cycle at 10 kHz sampling rate, which proves the feasibility of practical hardware deployment for real-time processing.

Ultrasensitive and Robust 120 keV Hard X-ray Imaging Detector Based on Mixed-Halide Perovskite CsPbBr3-n In Single Crystals.

Abstract: The relatively low resistivity and severe ion migration in inorganic perovskite CsPbBr3 significantly degrade the performance of X-ray detectors due to their high detection limit and current drift. We investigate the electrical properties and X-ray detection performances of CsPbBr3-n In single crystals by doping the iodine atoms into the melt-grown CsPbBr3 crystals. The resistivity of CsPbBr3-n In single crystals increases from 3.6 × 109 (CsPbBr3 ) to 2.2 × 1011 (CsPbBr2 I) Ω·cm as the iodine content increases, restraining the leak current and decreasing the detection limit of the detector. Additionally, CsPbBr3-n In single crystals exhibit stable dark currents under an electric field of 5000 V cm-1 , arising from their high ion migration activation energy. A record sensitivity of 6.3 × 104 μC Gy-1 cm-2 (CsPbBr2.9 I0.1 ) and a low detection limit of 54 nGy s-1 (CsPbBr2 I) are achieved by CsPbBr3-n In single crystals for the 120 keV hard X-ray detection under a 5000 V cm-1 electrical field. The CsPbBr2.9 I0.1 detector shows a stable current response with a dark current density of 0.58 μA cm-2 for 30 days and clear imaging for 120 keV X-rays at ambient conditions. The effective iodine atom doping strategy and the high quality make the CsPbBr3-n In SCs single crystals promising for reproducible high-energy hard X-ray imaging systems. This article is protected by copyright. All rights reserved.

Stochastic Gravitational-Wave Backgrounds: Current Detection Efforts and Future Prospects

Abstract: The collection of individually resolvable gravitational wave (GW) events makes up a tiny fraction of all GW signals that reach our detectors, while most lie below the confusion limit and are undetected. Similarly to voices in a crowded room, the collection of unresolved signals gives rise to a background that is well-described via stochastic variables and, hence, referred to as the stochastic GW background (SGWB). In this review, we provide an overview of stochastic GW signals and characterise them based on features of interest such as generation processes and observational properties. We then review the current detection strategies for stochastic backgrounds, offering a ready-to-use manual for stochastic GW searches in real data. In the process, we distinguish between interferometric measurements of GWs, either by ground-based or space-based laser interferometers, and timing-residuals analyses with pulsar timing arrays (PTAs). These detection methods have been applied to real data both by large GW collaborations and smaller research groups, and the most recent and instructive results are reported here. We close this review with an outlook on future observations with third generation detectors, space-based interferometers, and potential noninterferometric detection methods proposed in the literature.

Probing anisotropies of the Stochastic Gravitational Wave Background with LISA

Abstract: We investigate the sensitivity of the Laser Interferometer Space Antenna (LISA) to the anisotropies of the Stochastic Gravitational Wave Background (SGWB). We first discuss the main astrophysical and cosmological sources of SGWB which are characterized by anisotropies in the GW energy density, and we build a Signal-to-Noise estimator to quantify the sensitivity of LISA to different multipoles. We then perform a Fisher matrix analysis of the prospects of detectability of anisotropic features with LISA for individual multipoles, focusing on a SGWB with a power-law frequency profile. We compute the noise angular spectrum taking into account the specific scan strategy of the LISA detector. We analyze the case of the kinematic dipole and quadrupole generated by Doppler boosting an isotropic SGWB. We find that β ΩGW ∼ 2 × 10-11 is required to observe a dipolar signal with LISA. The detector response to the quadrupole has a factor ∼ 103 β relative to that of the dipole. The characterization of the anisotropies, both from a theoretical perspective and from a map-making point of view, allows us to extract information that can be used to understand the origin of the SGWB, and to discriminate among distinct superimposed SGWB sources.

Application of generative adversarial networks (GAN) for ophthalmology image domains: a survey

Abstract: Recent advances in deep learning techniques have led to improved diagnostic abilities in ophthalmology. A generative adversarial network (GAN), which consists of two competing types of deep neural networks, including a generator and a discriminator, has demonstrated remarkable performance in image synthesis and image-to-image translation. The adoption of GAN for medical imaging is increasing for image generation and translation, but it is not familiar to researchers in the field of ophthalmology. In this work, we present a literature review on the application of GAN in ophthalmology image domains to discuss important contributions and to identify potential future research directions.We performed a survey on studies using GAN published before June 2021 only, and we introduced various applications of GAN in ophthalmology image domains. The search identified 48 peer-reviewed papers in the final review. The type of GAN used in the analysis, task, imaging domain, and the outcome were collected to verify the usefulness of the GAN.In ophthalmology image domains, GAN can perform segmentation, data augmentation, denoising, domain transfer, super-resolution, post-intervention prediction, and feature extraction. GAN techniques have established an extension of datasets and modalities in ophthalmology. GAN has several limitations, such as mode collapse, spatial deformities, unintended changes, and the generation of high-frequency noises and artifacts of checkerboard patterns.The use of GAN has benefited the various tasks in ophthalmology image domains. Based on our observations, the adoption of GAN in ophthalmology is still in a very early stage of clinical validation compared with deep learning classification techniques because several problems need to be overcome for practical use. However, the proper selection of the GAN technique and statistical modeling of ocular imaging will greatly improve the performance of each image analysis. Finally, this survey would enable researchers to access the appropriate GAN technique to maximize the potential of ophthalmology datasets for deep learning research.

Flexible MXene/Bacterial Cellulose Film Sound Detector Based on Piezoresistive Sensing Mechanism.

Abstract: Flexible pressure sensors have aroused extensive attention in health monitoring, human-computer interaction, soft robotics, and more, as a staple member of wearable electronics. However, a majority of traditional research focuses solely on foundational mechanical sensing tests and ordinary human-motion monitoring, ignoring its other applications in daily life. In this work, a paper-based pressure sensor is prepared by using MXene/bacterial cellulose film with three-dimensional isolation layer structure, and its sensing capability as a wearable sound detector has also been studied. The as-prepared device exhibits great comprehensive mechanical sensing performance as well as accurate detection of human physiological signals. As a sound detector, not only can it recognize different voice signals and sound attributes by monitoring movement of throat muscles, but also it will distinguish a variety of natural sounds through air pressure waves caused by sound transmission (also called sound waves), like the eardrum. Besides, it plays an important role in sound visualization technology because of the ability for capturing and presenting music signals. Moreover, millimeter-scale thickness, lightweight, and degradable raw materials make the sensor convenient and easy to carry, meeting requirements of environmental protection as well.

Application of Neural Network and Time-Domain Feature Extraction Techniques for Determining Volumetric Percentages and the Type of Two Phase Flow Regimes Independent of Scale Layer Thickness

Abstract: One of the factors that significantly affects the efficiency of oil and gas industry equipment is the scales formed in the pipelines. In this innovative, non-invasive system, the inclusion of a dual-energy gamma source and two sodium iodide detectors was investigated with the help of artificial intelligence to determine the flow pattern and volume percentage in a two-phase flow by considering the thickness of the scale in the tested pipeline. In the proposed structure, a dual-energy gamma source consisting of barium-133 and cesium-137 isotopes emit photons, one detector recorded transmitted photons and a second detector recorded the scattered photons. After simulating the mentioned structure using Monte Carlo N-Particle (MCNP) code, time characteristics named 4th order moment, kurtosis and skewness were extracted from the recorded data of both the transmission detector (TD) and scattering detector (SD). These characteristics were considered as inputs of the multilayer perceptron (MLP) neural network. Two neural networks that were able to determine volume percentages with high accuracy, as well as classify all flow regimes correctly, were trained.

Miniaturized spectrometers with a tunable van der Waals junction

Abstract: Miniaturized computational spectrometers, which can obtain incident spectra using a combination of device spectral responses and reconstruction algorithms, are essential for on-chip and implantable applications. Highly sensitive spectral measurement using a single detector allows the footprints of such spectrometers to be scaled down while achieving spectral resolution approaching that of benchtop systems. We report a high-performance computational spectrometer based on a single van der Waals junction with an electrically tunable transport-mediated spectral response. We achieve high peak wavelength accuracy (∼0.36 nanometers), high spectral resolution (∼3 nanometers), broad operation bandwidth (from ∼405 to 845 nanometers), and proof-of-concept spectral imaging. Our approach provides a route toward ultraminiaturization and offers unprecedented performance in accuracy, resolution, and operation bandwidth for single-detector computational spectrometers. Description Miniaturizing spectrometers High-resolution spectrometry tends to be associated with bench-sized machines. Recent efforts on computational spectrometers have shown that this physical footprint can be shrunk by using nanowires and two-dimensional (2D) materials, but these devices are often associated with limited performance. Yoon et al. developed a single-detector computational spectrometer using an electrically tunable spectral response of a single junction comprising 2D van der Waal materials (see the Perspective by Quereda and Castellanos-Gomez). The electrically tunable spectral response and high performance of the tiny detector are promising for the further development of computational spectrometers. —ISO A single junction of two-dimensional van der Waal materials provides the basis for ultraminiaturized spectrometers.

Narrowband organic photodetectors - towards miniaturized, spectroscopic sensing.

Abstract: Omnipresent quality monitoring in food products, blood-oxygen measurement in lightweight conformal wrist bands, or data-driven automated industrial production: Innovation in many fields is being empowered by sensor technology. Specifically, organic photodetectors (OPDs) promise great advances due to their beneficial properties and low-cost production. Recent research has led to rapid improvement in all performance parameters of OPDs, which are now on-par or better than their inorganic counterparts, such as silicon or indium gallium arsenide photodetectors, in several aspects. In particular, it is possible to directly design OPDs for specific wavelengths. This makes expensive and bulky optical filters obsolete and allows for miniature detector devices. In this review, recent progress of such narrowband OPDs is systematically summarized covering all aspects from narrow-photo-absorbing materials to device architecture engineering. The recent challenges for narrowband OPDs, like achieving high responsivity, low dark current, high response speed, and good dynamic range are carefully addressed. Finally, application demonstrations covering broadband and narrowband OPDs are discussed. Importantly, several exciting research perspectives, which will stimulate further research on organic-semiconductor-based photodetectors, are pointed out at the very end of this review.

Memristor-Based Hyperchaotic Maps and Application in Auxiliary Classifier Generative Adversarial Nets

Abstract: With the nonlinearity and plasticity, memristors are widely used as nonlinear devices for chaotic oscillations or as biological synapses for neuromorphic computations. But discrete memristors (DMs) and their coupling maps have not received much attention, yet. Using a DM model, this article presents a general three-dimensional discrete memristor-based (3-D-DM) map model. By coupling the DM with four 2-D discrete maps, four examples of 3-D-DM maps with no or infinitely many fixed points are generated. We simulate the coupling coefficient-depended and memristor initial-boosted bifurcation behaviors of these 3-D-DM maps using numerical measures. The results demonstrate that the memristor can enhance the chaos complexity of existing discrete maps and its coupling maps can display hyperchaos. Furthermore, a hardware platform is developed to implement the 3-D-DM maps and the acquired hyperchaotic sequences have high randomness. Particularly, these hyperchaotic sequences can be applied to the auxiliary classifier generative adversarial nets for greatly improving the discriminator accuracy.

YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

Abstract: YOLOv7 surpasses all known object detectors in both speed and accuracy in the range from 5 FPS to 160 FPS and has the highest accuracy 56.8% AP among all known real-time object detectors with 30 FPS or higher on GPU V100. YOLOv7-E6 object detector (56 FPS V100, 55.9% AP) outperforms both transformer-based detector SWIN-L Cascade-Mask R-CNN (9.2 FPS A100, 53.9% AP) by 509% in speed and 2% in accuracy, and convolutional-based detector ConvNeXt-XL Cascade-Mask R-CNN (8.6 FPS A100, 55.2% AP) by 551% in speed and 0.7% AP in accuracy, as well as YOLOv7 outperforms: YOLOR, YOLOX, Scaled-YOLOv4, YOLOv5, DETR, Deformable DETR, DINO-5scale-R50, ViT-Adapter-B and many other object detectors in speed and accuracy. Moreover, we train YOLOv7 only on MS COCO dataset from scratch without using any other datasets or pre-trained weights. Source code is released in https://github.com/WongKinYiu/yolov7.

Miniaturized NIR Spectroscopy in Food Analysis and Quality Control: Promises, Challenges, and Perspectives

Abstract: The ongoing miniaturization of spectrometers creates a perfect synergy with the common advantages of near-infrared (NIR) spectroscopy, which together provide particularly significant benefits in the field of food analysis. The combination of portability and direct onsite application with high throughput and a noninvasive way of analysis is a decisive advantage in the food industry, which features a diverse production and supply chain. A miniaturized NIR analytical framework is readily applicable to combat various food safety risks, where compromised quality may result from an accidental or intentional (i.e., food fraud) origin. In this review, the characteristics of miniaturized NIR sensors are discussed in comparison to benchtop laboratory spectrometers regarding their performance, applicability, and optimization of methodology. Miniaturized NIR spectrometers remarkably increase the flexibility of analysis; however, various factors affect the performance of these devices in different analytical scenarios. Currently, it is a focused research direction to perform systematic evaluation studies of the accuracy and reliability of various miniaturized spectrometers that are based on different technologies; e.g., Fourier transform (FT)-NIR, micro-optoelectro-mechanical system (MOEMS)-based Hadamard mask, or linear variable filter (LVF) coupled with an array detector, among others. Progressing technology has been accompanied by innovative data-analysis methods integrated into the package of a micro-NIR analytical framework to improve its accuracy, reliability, and applicability. Advanced calibration methods (e.g., artificial neural networks (ANN) and nonlinear regression) directly improve the performance of miniaturized instruments in challenging analyses, and balance the accuracy of these instruments toward laboratory spectrometers. The quantum-mechanical simulation of NIR spectra reveals the wavenumber regions where the best-correlated spectral information resides and unveils the interactions of the target analyte with the surrounding matrix, ultimately enhancing the information gathered from the NIR spectra. A data-fusion framework offers a combination of spectral information from sensors that operate in different wavelength regions and enables parallelization of spectral pretreatments. This set of methods enables the intelligent design of future NIR analyses using miniaturized instruments, which is critically important for samples with a complex matrix typical of food raw material and shelf products.

PromptDet: Towards Open-Vocabulary Detection Using Uncurated Images

Abstract: The goal of this work is to establish a scalable pipeline for expanding an object detector towards novel/unseen categories, using zero manual annotations. To achieve that, we make the following four contributions: (i) in pursuit of generalisation, we propose a two-stage open-vocabulary object detector, where the class-agnostic object proposals are classified with a text encoder from pre-trained visual-language model; (ii) To pair the visual latent space (of RPN box proposals) with that of the pre-trained text encoder, we propose the idea of regional prompt learning to align the textual embedding space with regional visual object features; (iii) To scale up the learning procedure towards detecting a wider spectrum of objects, we exploit the available online resource via a novel self-training framework, which allows to train the proposed detector on a large corpus of noisy uncurated web images. Lastly, (iv) to evaluate our proposed detector, termed as PromptDet, we conduct extensive experiments on the challenging LVIS and MS-COCO dataset. PromptDet shows superior performance over existing approaches with fewer additional training images and zero manual annotations whatsoever. Project page with code: https://fcjian.github.io/promptdet .

Multipixel Anomaly Detection With Unknown Patterns for Hyperspectral Imagery

Abstract: In this article, anomaly detection is considered for hyperspectral imagery in the Gaussian background with an unknown covariance matrix. The anomaly to be detected occupies multiple pixels with an unknown pattern. Two adaptive detectors are proposed based on the generalized likelihood ratio test design procedure and ad hoc modification of it. Surprisingly, it turns out that the two proposed detectors are equivalent. Analytical expressions are derived for the probability of false alarm of the proposed detector, which exhibits a constant false alarm rate against the noise covariance matrix. Numerical examples using simulated data reveal how some system parameters (e.g., the background data size and pixel number) affect the performance of the proposed detector. Experiments are conducted on five real hyperspectral data sets, demonstrating that the proposed detector achieves better detection performance than its counterparts.

Learning to Match Anchors for Visual Object Detection

Abstract: Modern CNN-based object detectors assign anchors for ground-truth objects under the restriction of object-anchor Intersection-over-Union (IoU). In this study, we propose a learning-to-match (LTM) method to break IoU restriction, allowing objects to match anchors in a flexible manner. LTM updates hand-crafted anchor assignment to "free" anchor matching by formulating detector training in the Maximum Likelihood Estimation (MLE) framework. During the training phase, LTM is implemented by converting the detection likelihood to anchor matching loss functions which are plug-and-play. Minimizing the matching loss functions drives learning and selecting features which best explain a class of objects with respect to both classification and localization. LTM is extended from anchor-based detectors to anchor-free detectors, validating the general applicability of learnable object-feature matching mechanism for visual object detection. Experiments on MS COCO dataset demonstrate that LTM detectors consistently outperform counterpart detectors with significant margins. The last but not the least, LTM requires negligible computational cost in both training and inference phases as it does not involve any additional architecture or parameter. Code has been made publicly available.