Coarse to Fine: Domain Adaptive Crowd Counting via Adversarial Scoring Network

  title={Coarse to Fine: Domain Adaptive Crowd Counting via Adversarial Scoring Network},
  author={Zhikang Zou and Xiaoye Qu and Pan Zhou and Shuangjie Xu and Xiaoqing Ye and Wenhao Wu and Jinxing Ye},
  journal={Proceedings of the 29th ACM International Conference on Multimedia},
Recent deep networks have convincingly demonstrated high capability in crowd counting, which is a critical task attracting widespread attention due to its various industrial applications. Despite such progress, trained data-dependent models usually can not generalize well to unseen scenarios because of the inherent domain shift. To facilitate this issue, this paper proposes a novel adversarial scoring network (ASNet) to gradually bridge the gap across domains from coarse to fine granularity. In… 

Fine-Grained Fragment Diffusion for Cross Domain Crowd Counting

A cross- domain Fine-Grained Fragment Diffusion model (FGFD) is proposed that explores feature-level fine-grained similarities of crowd distributions between different fragments to bridge the cross-domain gap (content-level coarse- grained dissimilarities).

Forget Less, Count Better: A Domain-Incremental Self-Distillation Learning Benchmark for Lifelong Crowd Counting

A self-distillation learning framework is proposed as a benchmark for lifelong crowd counting, which helps the model sustainably leverage previous meaningful knowledge for better crowd counting to mitigate the forgetting when the new data arrive.

Learning Unbiased Transferability for Domain Adaptation by Uncertainty Modeling

This work dives into the transferability estimation problem in domain adaptation and proposes a non-intrusive Unbiased Transferability Estimation Plug-in (UTEP) by modeling the uncertainty of a discriminator in adversarial-based DA methods to optimize unbiased transfer.

SRNet: Scale-Aware Representation Learning Network for Dense Crowd Counting

The qualitative and quantitative results prove the effectiveness of the SRNet in dense crowd counting and crowd localization tasks and its role in improving the overall counting accuracy.

Backdoor Attacks on Crowd Counting

This paper proposes two novel Density Manipulation Backdoor Attacks (DMBA- and DMBA+) to attack the model to produce arbitrarily large or small density estimations, and provides an in-depth analysis of the unique challenges of backdooring crowd counting models.

Meta-Knowledge and Multi-Task Learning-Based Multi-Scene Adaptive Crowd Counting

A multi-scene adaptive crowd counting method based on meta-knowledge and multi-task learning that can be deployed to multiple new scenes without duplicated model training and reduces the MAE by 10.29% and 13.48%, respectively.

Rethinking Spatial Invariance of Convolutional Networks for Object Counting

Inspired by previous work, this work proposes a low-rank approximation accompanied with translation invariance to favorably implement the approximation of massive Gaussian convolution to improve the spatial invariance of convolutional networks.

Unsupervised Domain Adaptation for Semantic Segmentation of Urban Street Scenes Reflected by Convex Mirrors

This paper geometrically model convex mirrors to obtain a differentiable convex mirror simulation layer, CMSL, and performs adversarial domain adaptation on edges in the input space and semantic boundaries in the output space to reduce the geometric appearance gap between the synthetic and real images.

Video surveillance using deep transfer learning and deep domain adaptation: Towards better generalization



CODA: Counting Objects via Scale-Aware Adversarial Density Adaption

  • Li WangYongbo LiX. Xue
  • Computer Science
    2019 IEEE International Conference on Multimedia and Expo (ICME)
  • 2019
A novel adversarial learning approach to crowd counting, i.e., CODA (Counting Objects via scale-aware adversarial Density Adaption), which is comparable with the state-of-the-art fully-supervised models that are trained on the target dataset.

Leveraging Self-Supervision for Cross-Domain Crowd Counting

This work forces a network to learn perspective-aware features by training it to recognize upside-down real images from regular ones and incorporate into it the ability to predict its own uncertainty so that it can generate useful pseudo labels for fine-tuning purposes.

Discriminative Partial Domain Adversarial Network

In this paper, a novel Discriminative Partial Domain Adversarial Network (DPDAN) is developed and empirically verify DPDAN can effectively reduce the negative transfer caused by source-negative classes, and theoretically show it decreases negativeTransfer caused by domain shift.

Crowd Counting with Deep Negative Correlation Learning

This work proposes a new learning strategy to produce generalizable features by way of deep negative correlation learning (NCL), which deeply learn a pool of decorrelated regressors with sound generalization capabilities through managing their intrinsic diversities.

Crowd Counting With Deep Structured Scale Integration Network

A novel Deep Structured Scale Integration Network (DSSINet) for crowd counting is proposed, which addresses the scale variation of people by using structured feature representation learning and hierarchically structured loss function optimization.

Domain-adaptive Crowd Counting via Inter-domain Features Segregation and Gaussian-prior Reconstruction

A Domain-Adaptive Crowd Counting (DACC) framework, which consists of Inter-domain Features Segregation (IFS) and Gaussian-prior Reconstruction (GPR) and demonstrates that the proposed method outperforms the state-of-the-art methods.

Focus on Semantic Consistency for Cross-Domain Crowd Understanding

A domain adaptation method to eliminate a mass of estimation errors in the background areas of the synthetic and real-world crowd area and achieves the state-of-the-art for cross-domain counting problems.

DA-Net: Learning the Fine-Grained Density Distribution With Deformation Aggregation Network

The deformation aggregation network (DA-Net) is proposed that can incrementally incorporate adaptive receptive fields to capture the fine-grained density distribution and delivers the state-of-the-art performance on four benchmarks.

Feature-aware Adaptation and Structured Density Alignment for Crowd Counting in Video Surveillance

This paper proposes a domain-adaptation-style crowd counting method, which can effectively adapt the model from synthetic data to the specific real-world scenes, and outperforms the state-of-the-art methods for the same cross-domain counting problem.

Attentional Neural Fields for Crowd Counting

The CRFs coupled with the attention mechanism are seamlessly integrated into the encoder-decoder network, establishing an ANF that can be optimized end-to-end by back propagation, surpassing most previous methods.