Recently, counting the number of people for crowd scenes is a hot topic because its widespread applications (e.g. video surveillance, public security). It difficult task in wild: changeable environment, large-range cause current methods can not work well. In addition, due to scarce data, many suffer from over-fitting different extent. To remedy above two problems, firstly, we develop data collector and labeler, which generate synthetic simultaneously annotate them without any manpower. Based...

In the last decade, crowd counting and localization attract much attention of researchers due to its wide-spread applications, including monitoring, public safety, space design, etc. Many Convolutional Neural Networks (CNN) are designed for tackling this task. However, currently released datasets so small-scale that they can not meet needs supervised CNN-based algorithms. To remedy problem, we construct a large-scale congested dataset, NWPU-Crowd, consisting 5,109 images, in total 2,133,375...

Crowd counting from a single image is challenging task due to high appearance similarity, perspective changes, and severe congestion. Many methods only focus on the local features they cannot handle aforementioned challenges. In order tackle them, we propose crowd network (PCC Net), which consists of three parts: 1) density map estimation (DME) focuses learning very estimation; 2) random high-level classification (R-HDC) extracts global predict coarse labels patches in images; 3)...

Road detection from the perspective of moving vehicles is a challenging issue in autonomous driving. Recently, many deep learning methods spring up for this task because they can extract high-level local features to find road regions raw RGB data, such as Convolutional Neural Networks (CNN) and Fully (FCN). However, how detect boundary accurately still an intractable problem. In paper, we propose siamesed fully convolutional networks (named ``s-FCN-loc''), which able consider RGB-channel...

Semantic segmentation, a pixel-level vision task, is rapidly developed by using convolutional neural networks (CNNs). Training CNNs requires large amount of labeled data, but manually annotating data difficult. For emancipating manpower, in recent years, some synthetic datasets are released. However, they still different from real scenes, which causes that training model on the (source domain) cannot achieve good performance urban scenes (target domain). In this paper, we propose weakly...

Street scene understanding is an essential task for autonomous driving. One important step toward this direction labeling, which annotates each pixel in the images with a correct class label. Although many approaches have been developed, there are still some weak points. First, methods based on hand-crafted features whose image representation ability limited. Second, they cannot label foreground objects accurately due to data set bias. Third, refinement stage, traditional Markov random filed...

Accurately estimating the number of objects in a single image is challenging yet meaningful task and has been applied many applications such as urban planning public safety. In various object counting tasks, crowd particularly prominent due to its specific significance social security development. Fortunately, development techniques for can be generalized other related fields vehicle environment survey, if without taking their characteristics into account. Therefore, researchers are devoting...

With the development of deep neural networks, performance crowd counting and pixel-wise density estimation is continually being refreshed. Despite this, there are still two challenging problems in this field: 1) current supervised learning needs a large amount training data, but collecting annotating them difficult 2) existing methods cannot generalize well to unseen domain. A recently released synthetic dataset alleviates these problems. However, domain gap between real-world data images...

Cross-domain crowd counting (CDCC) is a hot topic due to its importance in public safety. The purpose of CDCC alleviate the domain shift between source and target domain. Recently, typical methods attempt extract domain-invariant features via image translation adversarial learning. When it comes specific tasks, we find that shifts are reflected on model parameters' differences. To describe gap directly at parameter-level, propose Neuron Linear Transformation (NLT) method, exploiting factor...

Recently, crowd counting using supervised learning achieves a remarkable improvement. Nevertheless, most counters rely on large amount of manually labeled data. With the release synthetic data, potential alternative is transferring knowledge from them to real data without any manual label. However, there no method effectively suppress domain gaps and output elaborate density maps during transferring. To remedy above problems, this article proposes domain-adaptive (DACC) framework, which...

Many CNN-based segmentation methods have been applied in lane marking detection recently and gain excellent success for a strong ability modeling semantic information. Although the accuracy of line prediction is getting better better, markings' localization relatively weak, especially when point remote. Traditional usually utilize highly specialized handcrafted features carefully designed postprocessing to detect lanes. However, these are based on assumptions and, thus, prone scalability. In...

For pixel-level crowd understanding, it is time-consuming and laborious in data collection annotation. Some domain adaptation algorithms try to liberate by training models with synthetic data, the results some recent works have proved feasibility. However, we found that a mass of estimation errors background areas impede performance existing methods. In this paper, propose method eliminate it. According semantic consistency, similar distribution deep layer's features real-world area, first...

Parkinson's disease (PD) is one of the most common neurodegenerative disorders aging, characterized by degeneration dopamine neurons (DA neurons) in substantial nigra, leading to advent both motor symptoms and non-motor symptoms. Current treatments include electrical stimulation affected brain areas replacement therapy. Even though categories are effective treating PD patients, progression cannot be stopped. The research advance into cell therapies provides exciting potential for treatment...

Road detection from the perspective of moving vehicles is a challenging issue in autonomous driving. Recently, many deep learning methods spring up for this task because they can extract high-level local features to find road regions raw RGB data, such as Convolutional Neural Networks (CNN) and Fully (FCN). However, how detect boundary accurately still an intractable problem. In paper, we propose siamesed fully convolutional network (named "s-FCN-loc") based on VGG-net architecture, which...

This technical report attempts to provide efficient and solid kits addressed on the field of crowd counting, which is denoted as Crowd Counting Code Framework (C$^3$F). The contributions C$^3$F are in three folds: 1) Some baseline networks presented, have achieved state-of-the-arts. 2) flexible parameter setting strategies provided further promote performance. 3) A powerful log system developed record experiment process, can enhance reproducibility each experiment. Our code made publicly...

Recently, counting the number of people for crowd scenes is a hot topic because its widespread applications (e.g. video surveillance, public security). It difficult task in wild: changeable environment, large-range cause current methods can not work well. In addition, due to scarce data, many suffer from over-fitting different extent. To remedy above two problems, firstly, we develop data collector and labeler, which generate synthetic simultaneously annotate them without any manpower. Based...

Heavy metal pollution in the soil results accumulation of heavy metals plants. While are toxic to plants, plants also resist toxicity through mechanism avoidance and tolerance. In this paper research progress on plant resistance stress is reviewed order provide scientific reference for further area.

Crowd localization is to predict each instance head position in crowd scenarios. Since the distance of pedestrians being camera are variant, there exists tremendous gaps among scales instances within an image, which called intrinsic scale shift. The core reason shift one most essential issues that it ubiquitous scenes and makes distribution chaotic. To this end, paper concentrates on access tackle chaos incurred by shift.We propose Gaussian Mixture Scope (GMS) regularize chaotic...