A real-time GPU implementation of the SIFT algorithm for large-scale video analysis tasks

Fassold, Hannes; Rosner, Jakub

doi:10.1117/12.2083201

Cited by 13 publications

(7 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One general way of speeding up processing is to make the feature extraction process for descriptors such as SIFT [2] more efficient. This can be done by more efficient and compact features (e.g., [3]), using GPUs (e.g., [4]) or dedicated hardware (e.g., [5], [6]) for feature extraction. As the feature extraction step for a single frame is treated as a black box in this paper, these approaches are complementary, and can be plugged into our framework.…”

Section: Related Workmentioning

confidence: 99%

Temporal Compression and Fast Matching of Hand-Crafted and Deep Features of Video Segments

Bailer

Wechtitsch

2018

2018 International Conference on Content-Based Multimedia Indexing (CBMI)

View full text Add to dashboard Cite

In order to enable efficient instance search in video, compact descriptors for video segments have been proposed. They exploit the temporal redundancy within a video segment to obtain smaller descriptors, and the segment-based representation can be exploited to enable more efficient matching. In this paper we analyze the performance of different visual features when applying both lossless and lossy compression to the set of descriptors of one video segment. We consider both handcrafted and deep features, i.e., visual features obtained from training a deep convolutional neural network. We also propose optimizations to the extraction and matching procedure. Both the compression methods and the optimizations are experimentally evaluated on a large video data set.

show abstract

Section: Related Workmentioning

confidence: 99%

Temporal Compression and Fast Matching of Hand-Crafted and Deep Features of Video Segments

Bailer

Wechtitsch

2018

2018 International Conference on Content-Based Multimedia Indexing (CBMI)

View full text Add to dashboard Cite

show abstract

“…A disadvantage of SIFT is the heavy computations required for the keypoints, where typical processing times are tenths of seconds to multiple seconds per frame in a normal CPU implementation. 24,25 Developments in graphics processing units…”

Section: Target Trackingmentioning

confidence: 99%

“…(GPUs) and field programmable gate arrays (FPGAs) have created opportunities for real-time algorithms. SIFT implementations have been developed for both GPUs [25][26][27] and FPGAs, 24 where the results demonstrate real-time SIFT calculations. SIFT features are composed of a keypoint that gives subpixel location and orientation of the feature, along with a descriptor that is calculated based on local pixel texture.…”

Section: Target Trackingmentioning

confidence: 99%

Multispectral detection and tracking of multiple moving targets in cluttered urban environments

2015

View full text Add to dashboard Cite

This paper presents an algorithm for target detection and tracking by fusion of multispectral imagery. In all spectral bands, we build a background model of the pixel intensities using a Gaussian mixture model, and pixels not belonging to the model are classified as foreground pixels. Foreground pixels from the spectral bands are weighted and summed into a single foreground map and filtered to give the fused foreground map. Foreground pixels are grouped into target candidates and associated with targets from a tracking database by matching features from the scale-invariant feature transform. The performance of our algorithm was evaluated with a synthetically generated data set of visible, near-infrared, midwave infrared, and long-wave infrared video sequences. With a fused combination of the spectral bands, the proposed algorithm lowers the false alarm rate while maintaining high detection rates. All 12 vehicles were tracked throughout the sequence, with one instance of a lost track that was later recovered.

show abstract

“…However, the current application domains require processing of images with resolutions ranging from 1280 × 720 (720 p) to 8192 × 4608 (8 k) known as highresolution images in real-time. Recent researches [16] [17] [18] [19] [20] have attempted to parallelize SIFT for high-resolution images. However, all of these attempts either have high execution time or fail to provide reproducible implementation, making current approaches of parallelizing the SIFT algorithm incompatible for real-time applications.…”

Section: Introductionmentioning

confidence: 99%

An Approach to Parallelization of SIFT Algorithm on GPUs for Real-Time Applications

Kumar¹,

Muknahallipatna²,

McInroy³

2016

JCC

View full text Add to dashboard Cite

Scale Invariant Feature Transform (SIFT) algorithm is a widely used computer vision algorithm that detects and extracts local feature descriptors from images. SIFT is computationally intensive, making it infeasible for single threaded implementation to extract local feature descriptors for high-resolution images in real time. In this paper, an approach to parallelization of the SIFT algorithm is demonstrated using NVIDIA's Graphics Processing Unit (GPU). The parallelization design for SIFT on GPUs is divided into two stages, a) Algorithm design-generic design strategies which focuses on data and b) Implementation design-architecture specific design strategies which focuses on optimally using GPU resources for maximum occupancy. Increasing memory latency hiding, eliminating branches and data blocking achieve a significant decrease in average computational time. Furthermore, it is observed via Paraver tools that our approach to parallelization while optimizing for maximum occupancy allows GPU to execute memory bound SIFT algorithm at optimal levels.

show abstract

A real-time GPU implementation of the SIFT algorithm for large-scale video analysis tasks

Cited by 13 publications

References 6 publications

Temporal Compression and Fast Matching of Hand-Crafted and Deep Features of Video Segments

Temporal Compression and Fast Matching of Hand-Crafted and Deep Features of Video Segments

Multispectral detection and tracking of multiple moving targets in cluttered urban environments

An Approach to Parallelization of SIFT Algorithm on GPUs for Real-Time Applications

Contact Info

Product

Resources

About