Heterogeneous CPU–GPU tracking–learning–detection (H-TLD) for real-time object tracking

Abstract

The recently proposed tracking-learning-detection (TLD) method has become a popular visual tracking algorithm as it was shown to provide promising longterm tracking results. On the other hand, the high computational cost of the algorithm prevents it being used at higher resolutions and frame rates. In this paper, we describe the design and implementation of a heterogeneous CPU-GPU TLD (H-TLD) solution using OpenMP and CUDA. Leveraging the advantages of the heterogeneous architecture, serial parts are run asynchronously on the CPU while the most computationally costly parts are parallelized and run on the GPU. Design of the solution ensures keeping data transfers between CPU and GPU at a minimum and applying stream compaction and overlapping data transfer with computation whenever such transfers are necessary. The workload is balanced for a uniform work distribution across the GPU multiprocessors. Results show that 10.25 times speed-up is achieved at 1920 Â 1080 resolution compared to the baseline TLD. The source code has been made publicly available to download from the following address: http://gpuresearch.ii. metu.edu.tr/codes/.

Efficient Parallel TLD on CPU-GPU Platform for Real-Time Tracking

“…The test dataset we adopted is OTB50 [3]. The baselines compared with our proposed parallel TLD include original OpenTLD, H-TLD [23] and AATLD [24]. Because this paper focuses on computing efficiency, we adopt execution time and speedup as the metric.…”

“…Recently, visual tracking has achieved a rapid development on accuracy and robustness rather than speed. There are few work discussing the parallel implementation and optimizations of trackers, especially on heterogeneous platform [21][22][23][24]. Research [23] proposes a high-performance version H-TLD based on OpenMP and CUDA.…”

“…There are few work discussing the parallel implementation and optimizations of trackers, especially on heterogeneous platform [21][22][23][24]. Research [23] proposes a high-performance version H-TLD based on OpenMP and CUDA. H-TLD exploits multicores on CPU and parallel units on GPU for load balance and decrease the transaction latency by data-compression techniques.…”

Fast Object Tracking on a Many-Core Neural Network Chip

“…However, these emerging neural network (NN) algorithms are usually very demanding in terms of compute and memory resources, limiting their execution speed. In addition, many of these algorithms usually involve several separate steps with heterogeneous operations to construct a complete tracking model (Gurcan and Temizel, 2015 ; Wang et al, 2017 ), which affects the hardware compatibility of all these different operations. To realize fast object tracking still remains as a challenge but important for applications such as motion posture capture in sports field (Chen et al, 2015 ; Pueo, 2016 ), cell imaging and movement analysis in biomedical field (Beier and Ibey, 2014 ), and some real-life scenarios (Galoogahi et al, 2017a ).…”