TILDE: A Temporally Invariant Learned DEtector

Verdie, Yannick; Yi, Kwang Moo; Fua, Pascal; Lepetit, Vincent

doi:10.1109/cvpr.2015.7299165

Cited by 287 publications

(267 citation statements)

References 42 publications

Supporting

Mentioning

261

Contrasting

Unclassified

Order By: Relevance

“…• Webcam dataset [40]: 6 sequences with 120 images in total. The dataset exhibits seasonal changes as well as daytime changes of scenes taken from far away.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Learning to Assign Orientations to Feature Points

Verdie

Fua

et al. 2016

2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

Self Cite

View full text Add to dashboard Cite

We show how to train a Convolutional Neural Network to assign a canonical orientation to feature points given an image patch centered on the feature point. Our method improves feature point matching upon the state-of-the art and can be used in conjunction with any existing rotation sensitive descriptors. To avoid the tedious and almost impossible task of finding a target orientation to learn, we propose to use Siamese networks which implicitly find the optimal orientations during training. We also propose a new type of activation function for Neural Networks that generalizes the popular ReLU, maxout, and PReLU activation functions. This novel activation performs better for our task. We validate the effectiveness of our method extensively with four existing datasets, including two non-planar datasets, as well as our own dataset. We show that we outperform the state-of-the-art without the need of retraining for each dataset.

show abstract

“…• Webcam dataset [40]: 6 sequences with 120 images in total. The dataset exhibits seasonal changes as well as daytime changes of scenes taken from far away.…”

Section: Resultsmentioning

confidence: 99%

“…To evaluate the performance of descriptors with orientations from the proposed method, we use datasets with both planar or far away objects [26,40,48] and 3D objects [1,36]. In addition, we created our own dataset as well, to further enrich the dataset with complex camera movements, such as in-plane rotations and viewpoint changes.…”

Section: Introductionmentioning

confidence: 99%

Learning to Assign Orientations to Feature Points

Verdie

Fua

et al. 2016

2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

Self Cite

View full text Add to dashboard Cite

show abstract

“…Soon after the introduction of deep learning it was also applied to the detection of key-points [14][15][16]. These methods are similar to handcrafted methods such as SIFT and ORB, but have the advantage that the networks can be trained to select key-points that are more apt for matching and image registration.…”

Section: Stable Region Detectormentioning

confidence: 99%

Stable Image Registration for In-Vivo Fetoscopic Panorama Reconstruction

et al. 2018

View full text Add to dashboard Cite

Abstract:A Twin-to-Twin Transfusion Syndrome (TTTS) is a condition that occurs in about 10% of pregnancies involving monochorionic twins. This complication can be treated with fetoscopic laser coagulation. The procedure could greatly benefit from panorama reconstruction to gain an overview of the placenta. In previous work we investigated which steps could improve the reconstruction performance for an in-vivo setting. In this work we improved this registration by proposing a stable region detection method as well as extracting matchable features based on a deep-learning approach. Finally, we extracted a measure for the image registration quality and the visibility condition. With experiments we show that the image registration performance is increased and more constant. Using these methods a system can be developed that supports the surgeon during the surgery, by giving feedback and providing a more complete overview of the placenta.

show abstract

“…They are often learning approaches: keypoint detection and matching are then trained from reference examples. Among these methods, TILDE (Verdié et al, 2015) formulates this problem as a regression model while LIFT (Yi et al, 2016) adopts deep learning models. The approach proposed in (Aubry et al, 2014) is adopted here.…”

Section: Automatic Ground Reference Identificationmentioning

confidence: 99%

Toward Automatic Georeferencing of Archival Aerial Photogrammetric Surveys

Giordano¹,

Bris²,

Mallet³

2018

ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

ABSTRACT:Images from archival aerial photogrammetric surveys are a unique and relatively unexplored means to chronicle 3D land-cover changes over the past 100 years. They provide a relatively dense temporal sampling of the territories with very high spatial resolution. Such time series image analysis is a mandatory baseline for a large variety of long-term environmental monitoring studies. The current bottleneck for accurate comparison between epochs is their fine georeferencing step. No fully automatic method has been proposed yet and existing studies are rather limited in terms of area and number of dates. State-of-the art shows that the major challenge is the identification of ground references: cartographic coordinates and their position in the archival images. This task is manually performed, and extremely time-consuming. This paper proposes to use a photogrammetric approach, and states that the 3D information that can be computed is the key to full automation. Its original idea lies in a 2-step approach: (i) the computation of a coarse absolute image orientation; (ii) the use of the coarse Digital Surface Model (DSM) information for automatic absolute image orientation. It only relies on a recent orthoimage+DSM, used as master reference for all epochs. The coarse orthoimage, compared with such a reference, allows the identification of dense ground references and the coarse DSM provides their position in the archival images. Results on two areas and 5 dates show that this method is compatible with long and dense archival aerial image series. Satisfactory planimetric and altimetric accuracies are reported, with variations depending on the ground sampling distance of the images and the location of the Ground Control Points.

show abstract

TILDE: A Temporally Invariant Learned DEtector

Cited by 287 publications

References 42 publications

Learning to Assign Orientations to Feature Points

Learning to Assign Orientations to Feature Points

Stable Image Registration for In-Vivo Fetoscopic Panorama Reconstruction

Toward Automatic Georeferencing of Archival Aerial Photogrammetric Surveys

Contact Info

Product

Resources

About