scholarly journals PARNet: A Joint Loss Function and Dynamic Weights Network for Pedestrian Semantic Attributes Recognition of Smart Surveillance Image

2019 ◽  
Vol 9 (10) ◽  
pp. 2027 ◽  
Author(s):  
Yong Li ◽  
Guofeng Tong ◽  
Xin Li ◽  
Yuebin Wang ◽  
Bo Zou ◽  
...  
Keyword(s):  
Loss Function ◽  
Practical Significance ◽  
Sigmoid Function ◽  
Recognition Method ◽  
Deep Model ◽  
Outdoor Scenes ◽  
Smart Surveillance ◽  
Attribute Recognition ◽  
Semantic Attributes ◽  
Indoor And Outdoor

The capability for recognizing pedestrian semantic attributes, such as gender, clothes color and other semantic attributes is of practical significance in bank smart surveillance, intelligent transportation and so on. In order to recognize the key multi attributes of pedestrians in indoor and outdoor scenes, this paper proposes a deep network with dynamic weights and joint loss function for pedestrian key attribute recognition. First, a new multi-label and multi-attribute pedestrian dataset, which is named NEU-dataset, is built. Second, we propose a new deep model based on DeepMAR model. The new network develops a loss function, which joins the sigmoid function and the softmax loss to solve the multi-label and multi-attribute problem. Furthermore, the dynamic weight in the loss function is adopted to solve the unbalanced samples problem. The experiment results show that the new attribute recognition method has good generalization performance.

Novel Scenes Improve Recollection and Recall of Words

2008 ◽  
Vol 20 (7) ◽  
pp. 1250-1265 ◽  
Author(s):  
Daniela B. Fenker ◽  
Julietta U. Frey ◽  
Hartmut Schuetze ◽  
Dorothee Heipertz ◽  
Hans-Jochen Heinze ◽  
...  
Keyword(s):  
Specific Activity ◽  
Activity Patterns ◽  
Long Term Potentiation ◽  
Contextual Memory ◽  
Memory Enhancement ◽  
Term Potentiation ◽  
Outdoor Scenes ◽  
Indoor And Outdoor ◽  
Hippocampus Dependent Memory

Exploring a novel environment can facilitate subsequent hippocampal long-term potentiation in animals. We report a related behavioral enhancement in humans. In two separate experiments, recollection and free recall, both measures of hippocampus-dependent memory formation, were enhanced for words studied after a 5-min exposure to unrelated novel as opposed to familiar images depicting indoor and outdoor scenes. With functional magnetic resonance imaging, the enhancement was predicted by specific activity patterns observed during novelty exposure in parahippocampal and dorsal prefrontal cortices, regions which are known to be linked to attentional orienting to novel stimuli and perceptual processing of scenes. Novelty was also associated with activation of the substantia nigra/ventral tegmental area of the midbrain and the hippocampus, but these activations did not correlate with contextual memory enhancement. These findings indicate remarkable parallels between contextual memory enhancement in humans and existing evidence regarding contextually enhanced hippocampal plasticity in animals. They provide specific behavioral clues to enhancing hippocampus-dependent memory in humans.

The Application of Entropy Weight of Attribute Recognition Model in Reservoir Eutrophication Evaluation

2013 ◽  
Vol 477-478 ◽  
pp. 870-873
Author(s):  
Du Wu ◽  
De Shan Tang ◽  
Xing Wang Lu ◽  
Wen Zhong Yu
Keyword(s):  
Weight Distribution ◽  
Guangdong Province ◽  
Element Model ◽  
Recognition Method ◽  
Entropy Weight ◽  
Recognition Model ◽  
Calculation Results ◽  
Attribute Recognition ◽  
Matter Element Model

Subjective factors could affect the weight distribution of each index in evaluation of reservoir eutrophication, so the example used entropy to deal with the weight distribution of each index. Combined attributes recognition method, the writer selected six indicators to build the entropy weight of attribute recognition model about reservoir eutrophication of ten large reservoirs in Guangdong Province. By comparing the calculated results with the results of matter-element model, the calculation results were basically consistent. So entropy weight of attribute recognition model is applicable to the evaluation of the reservoir eutrophication and it can ensure the fairness and reasonableness of weight distribution.

The Recognition of Handwritten Digital Based on BP Neural Network

2013 ◽  
Vol 416-417 ◽  
pp. 1239-1243
Author(s):  
Shan Gao
Keyword(s):  
Neural Network ◽  
Bp Neural Network ◽  
Recognition Rate ◽  
Practical Significance ◽  
Test Results ◽  
Recognition Method ◽  
Number Recognition ◽  
Recognition Ability ◽  
Fuzzy Recognition ◽  
Zip Code

The article put forward to new recognition method of handwritten digital based on BP neural network. Its recognition process mainly includes ten aspect: incline correction of handwritten number, edge detection and separation of a set number, binarization, denoising, extraction of numerals, window scaling, location standardization, thinning, extraction of numeral feature and fuzzy recognition based on BP neural network. The test results show that the recognition rate of this method can be over 92 percent. The recognition time of characters for character is less than 1.1 second, which means that the method is more effective recognition ability and can better satisfy the real system requirements.It should be widely applied practical significance for Book Number Recognition, zip code recognition sorting.

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

2019 ◽  
Vol 11 (4) ◽  
pp. 446 ◽  
Author(s):  
Zacharias Kandylakis ◽  
Konstantinos Vasili ◽  
Konstantinos Karantzalos
Keyword(s):  
Object Detection ◽  
Moving Objects ◽  
Local Coordinate System ◽  
Image Plane ◽  
Hyperspectral Data ◽  
Video Data ◽  
Outdoor Scenes ◽  
Indoor Scenes ◽  
Video Sensors ◽  
Indoor And Outdoor

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.

SEMANTIC VIDEO TRANSCODING USING CLASSES OF RELEVANCE

2003 ◽  
Vol 03 (01) ◽  
pp. 145-169 ◽  
Author(s):  
RITA CUCCHIARA ◽  
COSTANTINO GRANA ◽  
ANDREA PRATI
Keyword(s):  
Video Processing ◽  
Video Transcoding ◽  
User Requirements ◽  
User Interests ◽  
Object Based ◽  
Outdoor Scenes ◽  
Web Access ◽  
Evaluation Metric ◽  
Scene Object ◽  
Indoor And Outdoor

In this work we present a framework for on-the-fly video transcoding that exploits computer vision-based techniques to adapt the Web access to the user requirements. The proposed transcoding approach aims at coping with both user bandwidth and resources capabilities, and with user interests in the video's content. We propose an object-based semantic transcoding that, according to the user-defined classes of relevance, applies different transcoding techniques to the objects segmented in a scene. Object extraction is provided by on-the-fly video processing, without manual annotation. Multiple transcoding policies are reviewed and a performance evaluation metric based on the Weighted Mean Square Error (and corresponding PSNR), that takes into account the perceptual user requirements by means of classes of relevance, is defined. Results are analyzed by varying transcoding techniques, bandwidth requirements and video types (with indoor and outdoor scenes), showing that the use of semantics can dramatically improve the bandwidth to distortion ratio.

scholarly journals Signposts in the fog: objects facilitate scene representations in left scene-selective cortex

2017 ◽  
Author(s):  
Talia Brandman ◽  
Marius V. Peelen
Keyword(s):  
Visual Image ◽  
Object Perception ◽  
Neural Representation ◽  
Independent Experiment ◽  
Response Patterns ◽  
Outdoor Scenes ◽  
Scene Representation ◽  
Early Visual Cortex ◽  
Feature Overlap ◽  
Indoor And Outdoor

AbstractWe internally represent the structure of our surroundings even when there is little layout information available in the visual image, such as when walking through fog or darkness. One way in which we disambiguate such scenes is through object cues; for example, seeing a boat supports the inference that the foggy scene is a lake. Recent studies have investigated the neural mechanisms by which object and scene processing interact to support object perception. The current study examines the reverse interaction, by which objects facilitate the neural representation of scene layout. Photographs of indoor (closed) and outdoor (open) real-world scenes were blurred such that they were difficult to categorize on their own, but easily disambiguated by the inclusion of an object. fMRI decoding was used to measure scene representations in scene-selective parahippocampal place area (PPA) and occipital place area (OPA). Classifiers were trained to distinguish response patterns to fully visible indoor and outdoor scenes, presented in an independent experiment. Testing these classifiers on blurred scenes revealed a strong improvement in classification in left PPA and OPA when objects were present, despite the reduced low-level visual feature overlap with the training set in this condition. These findings were specific to left PPA/OPA, with no evidence for object-driven facilitation in right PPA/OPA, object-selective areas, and early visual cortex. These findings demonstrate separate roles for left and right scene-selective cortex in scene representation, whereby left PPA/OPA represents inferred scene layout, influenced by contextual object cues, and right PPA/OPA represents a scene’s visual features.

Sign in / Sign up

Export Citation Format

Share Document