Beyond Group: Multiple Person Tracking via Minimal Topology-Energy-Variation

Spatio-Temporal Point Process for Multiple Object Tracking

Multiple object tracking (MOT) focuses on modeling the relationship of detected objects among consecutive frames and merge them into different trajectories. MOT remains a challenging task as noisy and confusing detection results often hinder the final performance. Furthermore, most existing research are focusing on improving detection algorithms and association strategies. As such, we propose a novel framework that can effectively predict and mask-out the noisy and confusing detection results before associating the objects into trajectories. In particular, we formulate such "bad" detection results as a sequence of events and adopt the spatio-temporal point process to model such events. Traditionally, the occurrence rate in a point process is characterized by an explicitly defined intensity function, which depends on the prior knowledge of some specific tasks. Thus, designing a proper model is expensive and time-consuming, with also limited ability to generalize well. To tackle this problem, we adopt the convolutional recurrent neural network (conv-RNN) to instantiate the point process, where its intensity function is automatically modeled by the training data. Furthermore, we show that our method captures both temporal and spatial evolution, which is essential in modeling events for MOT. Experimental results demonstrate notable improvements in addressing noisy and confusing detection results in MOT data sets. An improved state-of-the-art performance is achieved by incorporating our baseline MOT algorithm with the spatio-temporal point process model.

Crowd Density Estimation in Spatial and Temporal Distortion Environment Using Parallel Multi-Size Receptive Fields and Stack Ensemble Meta-Learning

The estimation of crowd density is crucial for applications such as autonomous driving, visual surveillance, crowd control, public space planning, and warning visually distracted drivers prior to an accident. Having strong translational, reflective, and scale symmetry, models for estimating the density of a crowd yield an encouraging result. However, dynamic scenes with perspective distortions and rapidly changing spatial and temporal domains still present obstacles. The main reasons for this are the dynamic nature of a scene and the difficulty of representing and incorporating the feature space of objects of varying sizes into a prediction model. To overcome the aforementioned issues, this paper proposes a parallel multi-size receptive field units framework that leverages the majority of the CNN layer’s features, allowing for the representation and participation in the model prediction of the features of objects of all sizes. The proposed method utilizes features generated from lower to higher layers. As a result, different object scales can be handled at different framework depths, and various environmental densities can be estimated. However, the inclusion of the vast majority of layer features in the prediction model has a number of negative effects on the prediction’s outcome. Asymmetric non-local attention and the channel weighting module of a feature map are proposed to handle noise and background details and re-weight each channel to make it more sensitive to important features while ignoring irrelevant ones, respectively. While the output predictions of some layers have high bias and low variance, those of other layers have low bias and high variance. Using stack ensemble meta-learning, we combine individual predictions made with lower-layer features and higher-layer features to improve prediction while balancing the tradeoff between bias and variance. The UCF CC 50 dataset and the ShanghaiTech dataset have both been subjected to extensive testing. The results of the experiments indicate that the proposed method is effective for dense distributions and objects of various sizes.

Neuron Linear Transformation: Modeling the Domain Shift for Crowd Counting

Cross-domain crowd counting (CDCC) is a hot topic due to its importance in public safety. The purpose of CDCC is to alleviate the domain shift between the source and target domain. Recently, typical methods attempt to extract domain-invariant features via image translation and adversarial learning. When it comes to specific tasks, we find that the domain shifts are reflected in model parameters' differences. To describe the domain gap directly at the parameter level, we propose a neuron linear transformation (NLT) method, exploiting domain factor and bias weights to learn the domain shift. Specifically, for a specific neuron of a source model, NLT exploits few labeled target data to learn domain shift parameters. Finally, the target neuron is generated via a linear transformation. Extensive experiments and analysis on six real-world data sets validate that NLT achieves top performance compared with other domain adaptation methods. An ablation study also shows that the NLT is robust and more effective than supervised and fine-tune training. Code is available at https://github.com/taohan10200/NLT.

Bidirectional Interaction Network for Person Re-Identification

Person re-identification (ReID) task aims to retrieve the same person across multiple spatially disjoint camera views. Due to huge image changes caused by various factors such as posture variation and illumination transformation, images of different persons may share the more similar appearances than images of the same one. Learning discriminative representations to distinguish details of different persons is significant for person ReID. Many existing methods learn discriminative representations resorting to a human body part location branch which requires cumbersome expert human annotations or complex network designs. In this article, a novel bidirectional interaction network is proposed to explore discriminative representations for person ReID without any human body part detection. The proposed method regards multiple convolutional features as responses to various body part properties and exploits the inter-layer interaction to mine discriminative representations for person identities. Firstly, an inter-layer bilinear pooling strategy is proposed to feasibly exploit the pairwise feature relations between two convolution layers. Secondly, to explore interaction of multiple layers, an effective bidirectional integration strategy consisting of two different multi-layer interaction processes is designed to aggregate bilinear pooling interaction of multiple convolution layers. The interaction of multiple layers is implemented in a layer-by-layer nesting policy to ensure the two interaction processes are different and complementary. Extensive experiments validate the superiority of the proposed method on four popular person ReID datasets including Market-1501, DukeMTMC-ReID, CUHK03-NP and MSMT17. Specifically, the proposed method achieves a rank-1 accuracy of 95.1% and 88.2% on Market-1501 and DukeMTMC-ReID, respectively.

Fusing Multimodal Video Data for Detecting Moving Objects/Targets in Challenging Indoor and Outdoor Scenes

Single sensor systems and standard optical—usually RGB CCTV video cameras—fail to provide adequate observations, or the amount of spectral information required to build rich, expressive, discriminative features for object detection and tracking tasks in challenging outdoor and indoor scenes under various environmental/illumination conditions. Towards this direction, we have designed a multisensor system based on thermal, shortwave infrared, and hyperspectral video sensors and propose a processing pipeline able to perform in real-time object detection tasks despite the huge amount of the concurrently acquired video streams. In particular, in order to avoid the computationally intensive coregistration of the hyperspectral data with other imaging modalities, the initially detected targets are projected through a local coordinate system on the hypercube image plane. Regarding the object detection, a detector-agnostic procedure has been developed, integrating both unsupervised (background subtraction) and supervised (deep learning convolutional neural networks) techniques for validation purposes. The detected and verified targets are extracted through the fusion and data association steps based on temporal spectral signatures of both target and background. The quite promising experimental results in challenging indoor and outdoor scenes indicated the robust and efficient performance of the developed methodology under different conditions like fog, smoke, and illumination changes.