ABSTRACT
Machine learning algorithms are commonly used for quickly and efficiently counting people from a crowd. Test-time adaptation methods for crowd counting adjust model parameters and employ additional data augmentation to better adapt the model to the specific conditions encountered during testing. The majority of current studies concentrate on unsupervised domain adaptation. These approaches commonly perform hundreds of epochs of training iterations, requiring a sizable number of unannotated data of every new target domain apart from annotated data of the source domain. Unlike these methods, we propose a meta-test-time adaptive crowd counting approach called CrowdTTA, which integrates the concept of test-time adaptation into the meta-learning framework and makes it easier for the counting model to adapt to the unknown test distributions. To facilitate the reliable supervision signal at the pixel level, we introduce uncertainty by inserting the dropout layer into the counting model. The uncertainty is then used to generate valuable pseudo labels, serving as effective supervisory signals for adapting the model. In the context of meta-learning, one image can be regarded as one task for crowd counting. In each iteration, our approach is a dual-level optimization process. In the inner update, we employ a self-supervised consistency loss function to optimize the model so as to simulate the parameters update process that occurs during the test phase. In the outer update, we authentically update the parameters based on the image with ground truth, improving the model's performance and making the pseudo labels more accurate in the next iteration. At test time, the input image is used for adapting the model before testing the image. In comparison to various supervised learning and domain adaptation methods, our results via extensive experiments on diverse datasets showcase the general adaptive capability of our approach across datasets with varying crowd densities and scales.
Subject(s)
Machine Learning , Humans , Crowding , AlgorithmsABSTRACT
An efficient and more accurate millimeter-wave imaging algorithm, applied to a close-range monostatic personnel screening system, with consideration of dual path propagation loss, is presented in this paper. The algorithm is developed in accordance with a more rigorous physical model for the monostatic system. The physical model treats incident waves and scattered waves as spherical waves with a more rigorous amplitude term as per electromagnetic theory. As a result, the proposed method can achieve a better focusing effect for multiple targets in different range planes. Since the mathematical methods in classical algorithms, such as spherical wave decomposition and Weyl identity, cannot handle the corresponding mathematical model, the proposed algorithm is derived through the method of stationary phase (MSP). The algorithm has been validated by numerical simulations and laboratory experiments. Good performance in terms of computational efficiency and accuracy has been observed. The synthetic reconstruction results show that the proposed algorithm has significant advantages compared with the classical algorithms, and the reconstruction by using full-wave data generated by FEKO further verifies the validity of the proposed algorithm. Finally, the proposed algorithm performs as expected over real data acquired by our laboratory prototype.
ABSTRACT
Millimeter wave (MMW) technology is expanding rapidly into security screening for dangerous items concealed under clothing. It uses safe non-ionizing radiation and penetrates clothing well. We present a new planar system at Ka band for the three-dimensional simultaneous imaging of both sides of an inspected person where the images are produced in real time by a recently proposed generalized holographic reconstruction algorithm. Low-cost linear frequency modulation (LFM) radar technology is used along with a simple but efficient method for system calibration. Experimental characterization of the spatial resolution and the sensitivity of the system prototype has been carried out. It is established that the achieved spatial resolution is 6 mm or better if the item is not obscured by clothing and it may deteriorate to 7 mm depending on the clothing hiding the item. The spatial sensitivity is confirmed to be at least 2 mm.
ABSTRACT
A simple and fast single channel passive millimeter wave (PMMW) imaging system for public security check is presented in this paper. It distinguishes itself with traditional ones by an innovative scanning mechanism. Indoor experiments against human body with or without concealed items in clothes show that imaging could be completed in 3 s with angular resolution of about 0.7°. In addition, its field of view (FOV) is adjustable according to the size of actual target.
ABSTRACT
Metamaterial of dual-square array is proposed to design a dual-band circular polarizer. The novel design of asymmetric unit cell and layout of duplicate arrays significantly enhances the coupling between electric and magnetic fields. Simulation and measurement results show that the polarizer presents wide angle circular dichroism and circular birefringence. Moreover, the polarization conversion of the proposed metamaterial changes with frequency, incident angle, and polarization of incident waves. The fundamental mechanism behind is concluded to be the angle-dependent chirality and dispersion of our novel design.