Notes on the OpenSURF Library

By Christopher Evans. 2009

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

Remote Sensing Image Scene Classification: Benchmark and State of the Art (2017, methodology): “Except for sparse SIFT descriptor, there also exist dense SIFT that is computed in uniformly and densely sampled local regions and several extensions such as PCA-SIFT [114] and speed-up robust features (SURF) [115].”
DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition (2013, methodology): “DeCAF dramatically outperforms the baseline SURF feature available with the Ofﬁce dataset as well as the deep adaptive method of Chopra et al. (2013).”
Learning to See by Moving (2015, methodology): “Recent advances in computer vision have shown that visual features learnt by training neural networks for the task of object recognition using more than a million labelled images are useful for many computer vision tasks like semantic segmentation, object detection and action classification [22, 13,”
Unsupervised Learning of Visual Representations Using Videos (2015, methodology): “Given the trajectories of SURF interest points, we classify these points as moving if the flow magnitude is more than 0.5 pixels.”
Learning Spread-Out Local Feature Descriptors (2017, background): “One of the most popular hand-crafted descriptors is SIFT [10] and its variants [2, 24], which are widely used in the computer vision community.”
Learning Robust Representations by Projecting Superficial Statistics Out (2018, methodology): “In order to search the old computer vision techniques for a method that can extract more textural information and less semantic information, we experimented with three classifcal computer vision techniques: SURF (Bay et al., 2006), LBP (He & Wang, 1990), and GLCM (Haralick et al., 1973) on several d”
Interpretability Beyond Classification Output: Semantic Bottleneck Networks (2019, methodology): “Here, the authors compose a fixed set of independent object detectors based on simple feature descriptors like SIFT [18] and SURF [4] and perform the final classification task on the collection of all detector outputs.”
One-shot-learning gesture recognition using HOG-HOF features (2013, methodology): “This can be done, for example, using SIFT (Lowe, 1999), or SURF (Bay et al., 2006).”
Action recognition with trajectory-pooled deep-convolutional descriptors (2015, background): “They resort to SURF [2] feature matching and optical ﬂow based matching, as these two kinds of matching scheme are complementary to each other.”
Transitive Invariance for Self-Supervised Visual Representation Learning (2017, methodology): “As a brief introduction, this method first applies Improved Dense Trajectories (IDT) [53] on videos to extract SURF [2] feature points and their motion.”
Object-Centric Representation Learning from Unlabeled Videos (2016, methodology): “Speciﬁcally, they ﬁrst obtain SURF [33] interest points and use Improved Dense Trajectories (IDT) [34] to obtain the motion of each SURF point.”
Learning with Augmented Features for Heterogeneous Domain Adaptation (2012, methodology): “Speciﬁcally, SURF features (Bay et al., 2006) are extracted for all the images.”
Domain Adaptive Neural Networks for Object Recognition (2014, methodology): “The standard evaluation protocol on this data set is based on using the SURF feature descriptor (Bay et al., 2008) as inputs to the model.”
Deep Reconstruction-Classification Networks for Unsupervised Domain Adaptation (2016, neutral): “Furthermore, they usually operated on the SURF-based features [31] extracted from the raw pixels.”
Fracking Deep Convolutional Image Descriptors (2014, background): “In particular, in some of the tests we obtain up to a 169% increase in performance with SIFT as a baseline.”
The structure-from-motion reconstruction pipeline - a survey with focus on short image sequences (2010, methodology): “SIFT and SURF regard this by using repeated downsampling and increasing of the filter size, respectively .”
Cross-View Action Recognition via Transferable Dictionary Learning (2016, unknown): “2D SURF features [4], HOG features, SIFT features [30] were extended to 3D respectively in [18], [36], and [46].”
Evaluation of deep convolutional nets for document image classification and retrieval (2015, methodology): “As in previous work [14], the words were k-means clustered SURF features [3].”