53,940 Hits in 4.3 sec

Task-Driven Feature Pooling for Image Classification

Guo-Sen Xie, Xu-Yao Zhang, Xiangbo Shu, Shuicheng Yan, Cheng-Lin Liu
2015 2015 IEEE International Conference on Computer Vision (ICCV)  
Feature pooling is an important strategy to achieve high performance in image classification. However, most pooling methods are unsupervised and heuristic.  ...  In this paper, we propose a novel task-driven pooling (TDP) model to directly learn the pooled representation from data in a discriminative manner.  ...  strategy named task-driven pooling (TDP) for image classification.  ... 
doi:10.1109/iccv.2015.140 dblp:conf/iccv/XieZSYL15 fatcat:d4clvyocuveajk3h7dc64dcg3u

Multi-scale and Discriminative Part Detectors Based Features for Multi-label Image Classification

Gong Cheng, Decheng Gao, Yang Liu, Junwei Han
2018 Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence  
Convolutional neural networks (CNNs) have shown their promise for image classification task.  ...  maps with each feature map representing the occurrence probability of a particular part detector and learn DPD-based features by using a task-driven pooling scheme.  ...  By integrating feature learning into classification, we can use more information from   i O X to get a task-driven feature representation which is more suitable for classification than traditional pooling  ... 
doi:10.24963/ijcai.2018/90 dblp:conf/ijcai/ChengGLH18 fatcat:wqundh2u5nhrjfzaudaf476pai

Hierarchical task-driven feature learning for tumor histology

Heather D. Couture, J.S. Marron, Nancy E. Thomas, Charles M. Perou, Marc Niethammer
2015 2015 IEEE 12th International Symposium on Biomedical Imaging (ISBI)  
We explore this hierarchical and task-driven model in classifying malignant melanoma and the genetic subtype of breast tumors from histology images.  ...  Through learning small and large-scale image features, we can capture the local and architectural structure of tumor tissue from histology images.  ...  The patch-level classification results indicate that the task-driven method has great promise in learning subtle features that distinguish classes.  ... 
doi:10.1109/isbi.2015.7164039 dblp:conf/isbi/CoutureMTPN15 fatcat:hcjpjtoh2ndl3l62xhukeyye5i

Deep Convolutional Features for Image Based Retrieval and Scene Categorization [article]

Arsalan Mousavian, Jana Kosecka
2015 arXiv   pre-print
Instead of recognition, this paper focuses on the image retrieval problem and proposes a examines alternative pooling strategies derived for CNN features.  ...  We examine several pooling strategies and demonstrate superior performance on the image retrieval task (INRIA Holidays) at the fraction of the computational cost, while using a relatively small memory  ...  The work of [7] proposed a data driven method for computing the coarse geographical location of an image using simpler features like GIST and color histograms.  ... 
arXiv:1509.06033v1 fatcat:q6jj7jwbz5fotercxumrk53tsy

Indoor Scene Recognition using Task and Saliency-driven Feature Pooling

Marco Fornoni, Barbara Caputo
2012 Procedings of the British Machine Vision Conference 2012  
Our results prove this approach to be effective in the indoor scenario, while being also meaningful for other scene categorization tasks.  ...  This is usually done by combining multiple representations of different image regions, most often using a fixed 4 × 4, or pyramidal image-partitioning scheme.  ...  The classification is finally performed using SVM [3] , with C fixed to 100 for all the features and datasets.  ... 
doi:10.5244/c.26.98 dblp:conf/bmvc/FornoniC12 fatcat:cxuqczlng5b5toeaqokfxlqrsa

Using Deep Convolutional Neural Networks to Circumvent Morphological Feature Specification when Classifying Subvisible Protein Aggregates from Micro-Flow Images [article]

Christopher P. Calderon, Austin L. Daniels, Theodore W. Randolph
2017 arXiv   pre-print
of "morphological features" in a variety of tasks.  ...  We demonstrate that our new classifier (in combination with a sample "image pooling" strategy) can obtain nearly perfect predictions using as few as 20 FIM images from a given protein formulation in a  ...  In this work, we demonstrate a new strategy for using ConvNets for classification tasks.  ... 
arXiv:1709.00152v1 fatcat:g37dcyxxhvgnpbfv4mw6hqqu7y

MeshCNN: A Network with an Edge [article]

Rana Hanocka, Amir Hertz, Noa Fish, Raja Giryes, Shachar Fleishman and Daniel Cohen-Or
2019 arXiv   pre-print
We demonstrate the effectiveness of our task-driven pooling on various learning tasks applied to 3D meshes.  ...  MeshCNN learns which edges to collapse, thus forming a task-driven process where the network exposes and expands the important features while discarding the redundant ones.  ...  In particular, we use the edge-collapse technique [Hoppe 1997 ] for our task-driven pooling operator.  ... 
arXiv:1809.05910v2 fatcat:di5f53ex25hp3dgd7l7je4nlt4

Classifying Melanoma Skin Lesions Using Convolutional Spiking Neural Networks with Unsupervised STDP Learning Rule

Qian Zhou, Yan Shi, Zhenghua Xu, Ruowei Qu, Guizhi Xu
2020 IEEE Access  
We further propose to use feature selection to select more diagnostic features to improve the classification performance of our networks.  ...  Experimental results show that comparing to CNNs that need to be trained from scratch, our SNNs (with and without feature selection) not only achieve much better classification accuracies but also have  ...  The CNNs can learn from training image set and automatically extract important features for classification.  ... 
doi:10.1109/access.2020.2998098 fatcat:hnuwnpsuvvh55jjfrpjvvcsnnm

Neural encoding and interpretation for high-level visual cortices based on fMRI using image caption features [article]

Kai Qiao, Chi Zhang, Jian Chen, Linyuan Wang, Li Tong, Bin Yan
2020 arXiv   pre-print
However, previous studies almost used image features of those deep network models pre-trained on classification task to construct visual encoding models.  ...  In this study, we introduced one higher-level vision task: image caption (IC) task and proposed the visual encoding model based on IC features (ICFVEM) to encode voxels of high-level visual cortices.  ...  However, for high-level visual cortices, those deep network models driven by the image classification task seem to be relatively limited.  ... 
arXiv:2003.11797v1 fatcat:svhidctfyvf4dllmqskz3ybtlm

Knowledge-Driven Network for Object Detection

Yundong Wu, Jiajia Liao, Yujun Liu, Kaiming Ding, Shimin Li, Zhilin Zhang, Guorong Cai, Jinhe Su
2021 Algorithms  
In this paper, we present a knowledge-driven network (KDNet)—a new architecture that can aggregate and model keypoint relations to augment object features for detection.  ...  Object detection is a challenging computer vision task with numerous real-world applications.  ...  These novel modules can extract interesting features in the image for further classification and regression. Corner pooling layers.  ... 
doi:10.3390/a14070195 fatcat:cfnhqungzbgkzkmebab2a6dyny

Active Learning for Visual Question Answering: An Empirical Study [article]

Xiao Lin, Devi Parikh
2017 arXiv   pre-print
We present an empirical study of active learning for Visual Question Answering, where a deep VQA model selects informative question-image pairs from a pool and queries an oracle for answers to maximally  ...  and effective goal-driven active learning scoring function to pick question-image pairs for deep VQA models under the Bayesian Neural Network framework.  ...  Acknowledgements We thank Michael Cogswell and Qing Sun for discussions about the active learning strategies.  ... 
arXiv:1711.01732v1 fatcat:2mf3bfiljna67n5rv52ymbef4y

Target Driven Instance Detection [article]

Phil Ammirato, Cheng-Yang Fu, Mykhailo Shvets, Jana Kosecka, Alexander C. Berg
2019 arXiv   pre-print
We introduce a Target Driven Instance Detector(TDID), which modifies existing general object detectors for the instance recognition setting.  ...  For many applications, such as household robotics, a system may need to recognize a few very specific instances at a time.  ...  This pooled target feature vector is then both cross-correlated with, and subtracted from, every location in the scene image feature map.  ... 
arXiv:1803.04610v6 fatcat:6nr7yhdcpng3nfeyrhb66esnfm

Attentive Fashion Grammar Network for Fashion Landmark Detection and Clothing Category Classification

Wenguan Wang, Wenguan Wang, Yuanlu Xu, Jianbing Shen, Song-Chun Zhu
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
The latter is goal-driven, which directly enhances task-related features and can be learned in an implicit, top-down manner.  ...  For enhancing clothing category classification, our fashion network is encoded with two novel attention mechanisms, i.e., landmark-aware attention and category-driven attention.  ...  Thus the rest layers (pooling-4, conv5s, pooling-5, and fcs) of VGG-Net can be stacked for final cloth image classification.  ... 
doi:10.1109/cvpr.2018.00449 dblp:conf/cvpr/WangXSZ18 fatcat:6ab7jil4ffbhpik52cba5kcpae

A Comparative Analysis of Visual Encoding Models Based on Classification and Segmentation Task-Driven CNNs

Ziya Yu, Chi Zhang, Linyuan Wang, Li Tong, Bin Yan
2020 Computational and Mathematical Methods in Medicine  
Using functional magnetic resonance imaging (fMRI) data, the features of natural visual stimulation are extracted using a segmentation network (FCN32s) and a classification network (VGG16) with different  ...  However, the prediction performances of encoding models will have differences based on different networks driven by different tasks. Here, the impact of network tasks on encoding models is studied.  ...  In particular, task-driven deep networks performing different computer vision tasks can extract different features from the same image stimuli, resulting in variations in the performance of encoding models  ... 
doi:10.1155/2020/5408942 pmid:32802150 pmcid:PMC7416280 fatcat:l7v4hynddffwjgnvohbeh7cscu

Gradient Driven Learning for Pooling in Visual Pipeline Feature Extraction Models [article]

Derek Rose, Itamar Arel
2013 arXiv   pre-print
Hyper-parameter selection remains a daunting task when building a pattern recognition architecture which performs well, particularly in recently constructed visual pipeline models for feature extraction  ...  Preliminary results show moderate potential gains in classification accuracy and highlight areas of importance within the intermediate feature representation space.  ...  Open issues remain, particularly in the learning rate choice (for which we have traded the pooling structure for hyper-parameter η = 5e −5 ) and the number of maps needed to cover separate areas of invariance  ... 
arXiv:1301.3755v1 fatcat:ehfxjynfrjeovdfrjgispdvioy
« Previous Showing results 1 — 15 out of 53,940 results