Author manuscript, published in "Computer Vision and Pattern Recognition (2011)" Reconstructing an image from its local descriptors

2011

Citations in the corpus

• A Survey of Methods for Explaining Black Box Models (2018, background): “elated to producing a certain prediction by realizing a sort of articial image prototype. There is a signicant amount of work in understanding the representation of DNN by means of artifact images, [41,100,105,111]. We conclude this section presenting how [25] deals with artifacts in DNNs. Finding a”
• Visualizing Deep Convolutional Neural Networks Using Natural Pre-images (2015, methodology): “Weinzaepfel et al. [47] reconstructed an image from SIFT features using a large vocabulary of patches to invert individual detections and blended the results using Laplace (harmonic) interpolation.”
• HOGgles: Visualizing Object Detection Features (2013, background): “[22] were the first to reconstruct an image given its keypoint SIFT descriptors [13].”
• Using deep learning to model the hierarchical structure and function of a cell (2018, methodology): “A model trained to identify images of dogs might, upon later inspection, be seen to have neurons capturing interpretable properties like ‘tail’ or ‘furry’ 38–40 .”
• Modeling Image Patches with a Generic Dictionary of Mini-epitomes (2014, unknown): “However, the SIFT and other similar descriptors are not suitable for image generation tasks and are very difficult to visualize [28].”
• Understanding deep image representations by inverting them (2014, background): “An interesting conclusion of [33, 35] is that, while HOG and SIFT may not be exactly invertible, they capture a significant amount of information about the image.”
• Inverting Convolutional Networks with Convolutional Networks (2015, methodology): “There has been research on inverting various traditional computer vision representations: HOG [8], DSIFT [9], Local Binary Descriptors [10], Bag-of-Visual-Words [11].”
• Salient Deconvolutional Networks (2016, methodology): “Some recent ones such as [24] build on the idea of natural (regularized) pre-images introduced in [13], which in turn are based on prior contributions that applied pre-images to representations such as HOG [21], SIFT [22], BoVW [2,7], as well as early neural networks [23,10,9,12,6,20].”
• Learning visual biases from human imagination (2014, methodology): “Feature Visualization: Our work explores a novel application of feature visualizations [36, 35, 22].”
• Inverting Visual Representations with Convolutional Networks (2015, methodology): “[26] use the Harris affine keypoint detector which yields elliptic keypoints, and the number and the locations of the keypoints may be different from our case.”
• Understanding Neural Networks through Representation Erasure (2016, methodology): “…and visualizing neural models have also been significantly explored in vision (Vondrick et al., 2013; Vedaldi et al., 2014; Zeiler and Fergus, 2014; Weinzaepfel et al., 2011; Erhan et al., 2009; Simonyan et al., 2013; Klöppel et al., 2008), which we do not describe here for lack of space.”
• Image Reconstruction from Bag-of-Visual-Words (2014, methodology): “Among the several definitions of the objective function of QAP, a formulation by Lawler [52], defined as follows, is the most general form. min N∑ i,j,k,l=1 cijklxikxjl .”
• Privacy Preserving Image Queries for Camera Localization (2019, background): “[72] were the first to invert SIFT features.”
• Detecting Overfitting of Deep Generative Networks via Latent Recovery (2019, background): “Memorization and Privacy Beyond these aspects of memorization and practical evaluation of generators lies the important and debated issue of privacy: How to ensure that the data used for training cannot leak by some reverse engineering, such as CNN inversion [20] or reconstruction from feature [31] ”