Rethinking Attentive Object Detection via Neural Attention Learning.

Visual attention advances object detection by attending neural networks to object representations. While existing methods incorporate empirical modules to empower network attention, we rethink attentive object detection from the network learning perspective in this work. We propose a NEural Attention Learning approach (NEAL) which consists of two parts. During the back-propagation of each training iteration, we first calculate the partial derivatives (a.k.a. the accumulated gradients) of the classification output with respect to the input features. We refine these partial derivatives to obtain attention response maps whose elements reflect the contributions to the final network predictions. Then, we formulate the attention response maps as extra objective functions, which are combined together with the original detection loss to train detectors in an end-to-end manner. In this way, we succeed in learning an attentive CNN model without introducing additional network structures. We apply NEAL to the two-stage object detection frameworks, which are usually composed of a CNN feature backbone, a region proposal network (RPN), and a classifier. We show that the proposed NEAL not only helps the RPN attend to objects but also enables the classifier to pay more attention to the premier positive samples. To this end, the localization (proposal generation) and classification mutually benefit from each other in our proposed method. Extensive experiments on large-scale benchmark datasets, including MS COCO 2017 and Pascal VOC 2012, demonstrate that the proposed NEAL algorithm advances the two-stage object detector over state-of-the-art approaches.

CycleMLP: A MLP-Like Architecture for Dense Visual Predictions.

This article presents a simple yet effective multilayer perceptron (MLP) architecture, namely CycleMLP, which is a versatile neural backbone network capable of solving various tasks of dense visual predictions such as object detection, segmentation, and human pose estimation. Compared to recent advanced MLP architectures such as MLP-Mixer (Tolstikhin et al. 2021), ResMLP (Touvron et al. 2021), and gMLP (Liu et al. 2021), whose architectures are sensitive to image size and are infeasible in dense prediction tasks, CycleMLP has two appealing advantages: 1) CycleMLP can cope with various spatial sizes of images; 2) CycleMLP achieves linear computational complexity with respect to the image size by using local windows. In contrast, previous MLPs have O(N2) computational complexity due to their full connections in space. 3) The relationship between convolution, multi-head self-attention in Transformer, and CycleMLP are discussed through an intuitive theoretical analysis. We build a family of models that can surpass state-of-the-art MLP and Transformer models e.g., Swin Transformer (Liu et al. 2021), while using fewer parameters and FLOPs. CycleMLP expands the MLP-like models' applicability, making them versatile backbone networks that achieve competitive results on dense prediction tasks For example, CycleMLP-Tiny outperforms Swin-Tiny by 1.3% mIoU on ADE20 K dataset with fewer FLOPs. Moreover, CycleMLP also shows excellent zero-shot robustness on ImageNet-C dataset.

A torsional thrust balance with asymmetrical configuration for microthruster performance evaluation.

In this paper, a torsional thrust balance with an asymmetrical arm is designed and tested which is effective for the microthruster performance evaluation in the vacuum facilities with limited space. An optimization design method for the key parameters of the thrust balance has been developed. By utilizing the asymmetrical arm, a great resolution can be obtained with a restrained arm length. A novel printed circuit board electrostatic comb has been applied to the thrust balance calibration. Experimental results show that the comb is capable of producing steady force in the range of about 30 µN-3300 µN and an impulse bit of 7 µNs-777 µNs which can be further extended to nano-Newton second range with a shorter pulse width and a lower voltage. The calibration results show that the thrust balance has a great repeatability and reliability. The total uncertainty of the thrust stand is estimated to be 3.33% in the 1 µNs range.

Development and analysis of a novel printed circuit board electrostatic comb system for micro-newton thrust stand calibration.

An electrostatic calibration technique is highly flexible in producing a wide range of force and it is widely applied for nano-newton to micro-newton thrust stand calibration. This paper proposes a novel method for electrostatic comb implementation and related experiments have been carried out. Based on the printed circuit board and commercial fins, the comb can be realized flexibly with the output force conveniently extended. The force generated by this kind of comb is theoretically analyzed. Different from the traditional comb structure, the conductive area of the comb fixed plate is minimized to improve the force consistency over engagement. The influence of fin length, fin number, applied voltage, and engagement on the output force has been studied experimentally. The final comb system is capable of producing steady force in the range 13-5040 µN with the relative error within 5%. With a high voltage pulse generator, this system could produce calibration impulse bit in the range 1-1000 µN s for which the lowest level can be far more extended to the nanonewton range with a shorter pulse width, a lower voltage, and a reduced number of fins. Moreover, the calibrator has a rather flat force-engagement characteristic when the engagement is in the range of 6 mm-16 mm, while the variation of electrostatic force is within 5%. This calibrator has a wide output range and great consistency, and it is beneficial for the thrust stand calibration.