RGB-T crowd counting (RGB-T CC) aims to estimate the crowd population size utilizing the complementary information from visible and thermal images . Current deep models for RGB-T CC typically adopt a three-tier architecture, featuring a middle fusion layer that aggregates both RGB and thermal streams. However, we find that this dedicated fusion layer dominates the training process, causing under-optimization of both modal branches, which becomes the performance bottleneck in mainstream multi-modal counting models. To address this challenge, we propose a simple-yet-effective counting architecture, the Spatial Exchanging Fusion Network (SEFNet). It is built on a Dual Attention Guided Spatial Exchanging (DASE) mechanism, enabling direct extraction and exchange of modality-complementary features between modalities without the extra fusion branch employed in most existing works. This design ensures a more balanced gradient back-propagation over networks, attaining optimized representations in multi-modality fusion over prior models. Besides, the Modality Gradient Enhancement Module (MGEM) in SEFNet can effectively learn modality-specific crowd representations with two counting sub-tasks, dynamically achieving better gradient distribution and further enhancing optimization in both modalities. Extensive experiments demonstrate that SEFNet significantly outperforms state-of-the-art methods on mainstream benchmark datasets, and also exhibits promising generalization ability across various counting backbones and losses.

RGB-Infrared person re-identification (RGB-IR ReID) aims to associate people across disjoint RGB and IR camera views. Currently, state-of-the-art performance of RGB-IR ReID is not as impressive as that of conventional ReID. Much of that is due to the notorious modality bias training issue brought by the single-modality ImageNet pre-training, which might yield RGB-biased representations that severely hinder the cross-modality image retrieval. This paper makes first attempt to tackle the task from a pre-training perspective. We propose a self-supervised pre-training solution, named Modality-Aware Multiple Granularity Learning (MMGL), which directly trains models from scratch only on multi-modal ReID datasets, but achieving competitive results against ImageNet pre-training, without using any external data or sophisticated tuning tricks. First, we develop a simple-but-effective ‘permutation recovery’ pretext task that globally maps shuffled RGB-IR images into a shared latent permutation space, providing modality-invariant global representations for downstream ReID tasks. Second, we present a part-aware cycle-contrastive (PCC) learning strategy that utilizes cross-modality cycle-consistency to maximize agreement between semantically similar RGB-IR image patches. This enables contrastive learning for the unpaired multi-modal scenarios, further improving the discriminability of local features without laborious instance augmentation. Based on these designs, MMGL effectively alleviates the modality bias training problem. Extensive experiments demonstrate that it learns better representations (+8.03% Rank-1 accuracy) with faster training speed (converge only in few hours) and higher data efficiency (<5% data size) than ImageNet pre-training. The results also suggest it generalizes well to various existing models, losses and has promising transferability across datasets. The code will be released at https://github.com/hansonchen1996/MMGL.

RGB-Infrared person re-identification (RGB-IR ReID) is a challenging cross-modality retrieval problem, which aims at matching the person-of-interest over visible and infrared camera views. Most existing works achieve performance gains through manually-designed feature selection modules, which often require significant domain knowledge and rich experience. In this paper, we study a general paradigm, termed Neural Feature Search (NFS), to automate the process of feature selection. Specifically, NFS combines a dual-level feature search space and a differentiable search strategy to jointly select identity-related cues in coarse-grained channels and fine-grained spatial pixels. This combination allows NFS to adaptively filter background noises and concentrate on informative parts of human bodies in a data-driven manner. Moreover, a cross-modality contrastive optimization scheme further guides NFS to search features that can minimize modality discrepancy whilst maximizing inter-class distance. Extensive experiments on mainstream benchmarks demonstrate that our method outperforms state-of-the-arts, especially achieving better performance on the RegDB dataset with significant improvement of 11.20% and 8.64% in Rank-1 and mAP, respectively.

Geo-tagged travel photos on social networks often contain location data such as points of interest (POIs), and also users’ travel preferences. In this paper, we propose a hybrid ensemble learning method, BAyes-Knn, that predicts personalized tourist routes for travelers by mining their geographical preferences from these location-tagged data. Our method trains two types of base classifiers to jointly predict the next travel destination: (1) The K-nearest neighbor (KNN) classifier quantifies users’ location history, weather condition, temperature and seasonality and uses a feature-weighted distance model to predict a user’s personalized interests in an unvisited location. (2) A Bayes classifier introduces a smooth kernel function to estimate a-priori probabilities of features and then combines these probabilities to predict a user’s latent interests in a location. All the outcomes from these subclassifiers are merged into one final prediction result by using the Borda count voting method. We evaluated our method on geo-tagged Flickr photos and Beijing weather data collected from 1 January 2005 to 1 July 2016. The results demonstrated that our ensemble approach outperformed 12 other baseline models. In addition, the results showed that our framework has better prediction accuracy than do context-aware significant travel-sequence-patterns recommendations and frequent travel-sequence patterns.

The relief models made by the common modeling software are difficult to be consistent with the digital terrain model (DTM) and cannot fully express their morphological characteristics (such as the form, the structure, the spatial distributions and so on). In view of the above question, this paper advances an automatic modeling method of 3D relief symbols with building-up style. A component library of 3D relief symbols will be built, which consists of three parts: the public component library, the private component library and the metadata database. Some components of the different 3D relief symbols will be encapsulated and put into the public component library if their models can be built by the similar or the same method during the modeling process, and the others will be encapsulated and put into the private component library. Each component in the public or private component library will be described by a metadata in the metadata database. All the models of 3D relief symbols can be built by some components in the component library. The method can be used to reduce the production cost and improve the production efficiency of 3D relief symbol.