Based on analyzing the character of cascaded decoder architecture commonly adopted in existing DETR-like models, this paper proposes a new decoder architecture. The cascaded decoder architecture constrains object queries to update in the cascaded direction, only enabling object queries to learn relatively-limited information from image features. However, the challenges for object detection in natural scenes (e.g., extremely-small, heavily-occluded, and confusingly mixed with the background) require an object detection model to fully utilize image features, which motivates us to propose a new decoder architecture with the parallel Multi-time Inquiries (MI) mechanism. MI mechanism is very simple, enabling object queries to parallelly perform multi-time inquiries to learn more comprehensive information from image features. Our MI based model, MI-DETR, outperforms all existing DETR-like models on COCO benchmark under different backbones and training epochs, achieving +2.3 AP and +0.6 AP improvements compared to the most representative model DINO and SOTA model Relation-DETR under ResNet-50 backbone.

This paper is motivated by an interesting phenomenon: the performance of object detection lags behind that of instance segmentation (i.e., performance imbalance) when investigating the intermediate results from the beginning transformer decoder layer of MaskDINO (i.e., the SOTA model for joint detection and segmentation). This phenomenon inspires us to think about a question: will the performance imbalance at the beginning layer of transformer decoder constrain the upper bound of the final performance? With this question in mind, we further conduct qualitative and quantitative pre-experiments, which validate the negative impact of detection-segmentation imbalance issue on the model performance. To address this issue, this paper proposes DI-MaskDINO model, the core idea of which is to improve the final performance by alleviating the detection-segmentation imbalance. DI-MaskDINO is implemented by configuring our proposed De-Imbalance (DI) module and Balance-Aware Tokens Optimization (BATO) module to MaskDINO. DI is responsible for generating balance-aware query, and BATO uses the balance-aware query to guide the optimization of the initial feature tokens. The balance-aware query and optimized feature tokens are respectively taken as the Query and Key&Value of transformer decoder to perform joint object detection and instance segmentation. DI-MaskDINO outperforms existing joint object detection and instance segmentation models on COCO and BDD100K benchmarks, achieving +1.2 $AP^{box}$ and +0.9 $AP^{mask}$ improvements compared to SOTA joint detection and segmentation model MaskDINO. In addition, DI-MaskDINO also obtains +1.0 $AP^{box}$ improvement compared to SOTA object detection model DINO and +3.0 $AP^{mask}$ improvement compared to SOTA segmentation model Mask2Former.

Fingerprint matching localization based on indoor ubiquitous wireless signal have been widely used in indoor localization. However, it’s labor-intensive to collect and update the radio map which consists of fingerprints, while crowdsourcing might be a potential way to solve the problem due to the popularization of smartphones. The radio map construction with crowdsourcing can be regarded as a topology construction problem, and manifold learning is popularly used to construct the topology from high-dimensional fingerprint space to two-dimensional plane. The challenge is how to obtain accurate location labels and construct the radio map from crowdsourcing data without any position information. In this paper, the performance of three typical manifold learning algorithms was compared, including the effectiveness of topology construction, the robustness to noise and the preservation of global and local structures. A comprehensive evaluation method was used to find the optimal topology construction algorithm and corresponding parameters for different scenes. Experimental results show that LLE and Isomap are suitable for simple indoor scenes and t-SNE is much more robust for complex scenes.

Recent years have witnessed a growing list of systems for distributed data-parallel training. Existing systems largely fit into two paradigms, i.e., parameter server and MPI-style collective operations. On the algorithmic side, researchers have proposed a wide range of techniques to lower the communication via "system relaxations": quantization, decentralization, and communication delay. However, most, if not all, existing systems only rely on standard synchronous and asynchronous stochastic gradient (SG) based optimization, therefore, cannot take advantage of all possible optimizations that the machine learning community has been developing recently. Given this emerging gap between the current landscapes of systems and theory, we build Bagua, a MPI-style communication library, providing a collection of primitives, that is both flexible and modular to support state-of-the-art system relaxation techniques of distributed training. Powered by this design, Bagua has a great ability to implement and extend various state-of-the-art distributed learning algorithms. In a production cluster with up to 16 machines (128 GPUs), Bagua can outperform PyTorch-DDP, Horovod and BytePS in the end-to-end training time by a significant margin (up to 2X) across a diverse range of tasks. Moreover, we conduct a rigorous tradeoff exploration showing that different algorithms and system relaxations achieve the best performance over different network conditions.

In order to overcome the problems of traditional link network sensitive data anti tampering operation, such as long time-consuming and low data security, a tamper proof model of link network sensitive data based on blockchain technology is proposed. Calculate the evenly distributed random variables of sensitive node data and the difference of running distance to obtain the probability of meeting the sensitive data with other neighbor nodes, and determine the sensitive data in the link network; obtain the frequency domain of the sensitive data of the infected link network through the square difference function, and calculate the membership mean value of the infected data samples in the sensitive data; analyze the working principle of blockchain technology, Set the master key and public key of sensitive data, generate the encryption key of sensitive data of link network, and use blockchain technology to complete the design of tamper proof model of sensitive data in link network. The experimental results show that the shortest time-consuming of the proposed method is about 1 s, and the maximum tamper proof security factor is about 9.7.

A foot-mounted pedestrian dead reckoning system is a self-contained technique for indoor localization. An inertial pedestrian navigation system includes wearable MEMS inertial sensors, such as an accelerometer, gyroscope, barometer, or magnetometer, which enable the measurement of the step length and the heading direction. In this plan, a method based on IMU/EKF+HMM+ZUPT+ZARU+HDR+the Earth Magnetic Yaw was designed to realize foot-mounted pedestrian navigation. Based on the characteristics of pedestrian navigation, the general likelihood ratio test (GLRT) and the Hidden Markov Model (HMM) were used to realize the detection of zero speed interval at different speed states. When the zero speed state is detected, the zero velocity update (ZUPT) method is used to limit the accumulation of IMU. The Zero Angular Rate Update (ZARU) + (heuristic heading reduction) HDR+the Earth Magnetic Yaw method is used to limit the IMU attitude and heading drift. Finally, the EKF method is used to realize the effective estimation and feedback of the speed, attitude and heading error of the pedestrian navigation system. Meanwhile, a fault detection algorithm based on the innovation vector is added to the EKF system to effectively detect and eliminate the gross errors in the measurements, to improve the filtering effect of EKF algorithm, and ensure the accuracy of pedestrian navigation results.