Multi-scale pyramid pooling for deep convolutional representation

By Donggeun Yoo, Sunggyun Park, Joon-Young Lee, In-So Kweon. 2015

102 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Growing a Brain: Fine-Tuning by Increasing Model Capacity (2017, neutral): “M [32] 86.8 CNNaug-SVM [32] 61.8 — — Task Customized CNNs Caffe-DAG [53] 64.6 LSVM [30] 87.1 LSVM [30] 61.4 Deep Optimized [4] 66.4 — — MsML+ [30] 89.5 DeCaf+DPD [7] 65.0 — — Places-CNN [59] 68.2 MPP [55] 91.3 MsML+ [30] 66.6 — — — — Deep Optimized [4] 91.3 MsML+* [30] 67.9 — — Data Augmented CNNs C”
Pixel-Level Domain Transfer (2016, background): “pre-train a large network [22] over ImageNet [23] and transfer the parameters to a target domain [12,24,25].”
P2T: Pyramid Pooling Transformer for Scene Understanding (2021, methodology): “After that, pyramid pooling has been widely applied to various vision tasks such as semantic segmentation [37], [54], [55], [56], [57] and object detection [36], [58], [59], [60].”
The History Began from AlexNet: A Comprehensive Survey on Deep Learning Approaches (2018, background): “The multi-scale pyramid pooling was proposed in 2015 [118].”
SIFT Meets CNN: A Decade Survey of Instance Retrieval (2016, methodology): “ure of the features to be fused. In the multi-scale fusion strategy, image patches of multiple scales are generated, from which the CNN features are extracted and subsequently pooled [26]. Yoo et al. [71] use Fisher Vector in image classiﬁcation to encode the FC layers extracted from multiple image ”
Deep learning for visual understanding: A review (2016, methodology): “Although FCN is proposed mainly for semantic segmentation, the technique could also be utilized in other applications, e.g. image classi ﬁ cation [191], edge detection [188] etc.”
Scene Recognition with CNNs: Objects, Scales and Dataset Bias (2016, methodology): “In contrast to recent works[4, 3, 21, 1], we adopt simple max pooling to aggregate patch features into image features.”
Deep learning for visual understanding (2017, unknown): “Second, in contrast to the features encoded by the variants of BoW [70, 230], BoSP is relatively in low dimension, making it advantageous in processing large scale datasets.”
Good Practice in CNN Feature Transfer (2016, methodology): “Yoo et al [43] propose to encode dense activations of the FC layers with Fisher Vector [28] from multiple image scales for image classiﬁcation.”