Dense Point Trajectories by GPU-Accelerated Large Displacement Optical Flow

By Narayanan Sundaram, Thomas Brox, Kurt Keutzer. 2010

526 citations recorded by the source corpus.

Subfields: Aerospace Engineering, 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 Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks (2017, background): “In visual tracking, enforcing simple forward-backward consistency has been a standard trick for decades [44].”
FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks (2016, background): “Image Overlay FlowFields [2] DeepFlow [31] LDOF (GPU) [26] PCA-Flow [32] FlowNetS [10] FlowNet2”
Image Inpainting for Irregular Holes Using Partial Convolutions (2018, positive): “We found the results of occlusion/dis-occlusion mask estimation method between two consecutive frames for videos described in [29] to be a good source of such patterns.”
Visual Question Generation as Dual Task of Visual Question Answering (2017, methodology): “It has been used as a standard trick for years in visual tracking to enforce forward-backward consistency [29].”
Cycle-Consistency for Robust Visual Question Answering (2019, neutral): “Using cycle-consistency to regularize the training of models has been used extensively in object tracking [40], machine translation [11], unpaired image-to-image translation [48] and text-based question answering [41].”
Optical Flow Estimation Using a Spatial Pyramid Network (2016, methodology): “Using a GPU can speed up traditional methods [37, 43] but with reduced accuracy.”
Drag Your GAN: Interactive Point-based Manipulation on the Generative Image Manifold (2023, methodology): “Conventional approaches solve optimization problems with hand-crafted criteria [Brox and Malik 2010; Sundaram et al. 2010], while deep learning-based approaches started to dominate the field in recent years due to better performance [Dosovitskiy et al.”
Object Segmentation by Long Term Analysis of Point Trajectories (2010, methodology): “In order to compute such trajectories, we run a tracker we developed in [2], which is based on large displacement optical flow [3].”
Spatio-Temporal Moving Object Proposals (2014, methodology): “Works of [8,31] cluster dense point trajectories [36] using 2D motion similarities.”
DDFlow: Learning Optical Flow with Unlabeled Data Distillation (2019, methodology): “Our occlusion detection is based on the forward-backward consistency prior (Sundaram, Brox, and Keutzer 2010; Meister, Hur, and Roth 2018).”
Coherent Online Video Style Transfer (2017, methodology): “For the ground-truth of the composition mask M , we adopt the methods used in [37, 39] to detect occlusions and motion boundaries.”
Learning Motion Patterns in Videos (2016, methodology): “The overall method ‘MP-Net + Objectness + CRF’ performs better with LDOF (69.7) than EpicFlow (68.0).”
SelFlow: Self-Supervised Learning of Optical Flow (2019, methodology): “For two-frame optical flow estimation, we can swap two images as input to generate forward and backward flow, then the occlusion map can be generated based on the forward-backward consistency prior [44, 29].”
Video Segmentation with Superpixels (2012, background): “In order to calculate superpixel affinities across many frames that are potentially hundreds of frames apart we leverage on the recently introduced long-term point trajectories of [29].”
Artistic Style Transfer for Videos and Spherical Images (2017, methodology): “To detect disocclusions, we perform a forward-backward consistency check of the optical flow (Sundaram et al. 2010).”
Semantic Instance Annotation of Street Scenes by 3D to 2D Label Transfer (2015, methodology): “+ GC”) using the feature tracking approach of [41] to propagate semantic labels from the two closest labeled frames to the target frame.”
Efficient sparse-to-dense optical flow estimation using a learned basis and layers (2015, methodology): “[47] port the large-displacement optical flow method [9] to a GPU, with a runtime of 1.”
UnFlow: Unsupervised Learning of Optical Flow with a Bidirectional Census Loss (2017, methodology): “Our occlusion detection is based on the forward-backward consistency assumption (Sundaram, Brox, and Keutzer 2010).”
Action recognition by dense trajectories (2011, methodology): “Once the dense optical ﬂow ﬁeld is computed, points can be tracked very densely without additional cost.”