Structured Model Pruning of Convolutional Networks on Tensor Processing Units

Chen, Kongtao; Franko, Ken; Sang, Ruoxin

doi:10.48550/arxiv.2107.04191

Cited by 11 publications

(9 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the future, we will design loss functions that can better reflect the quality difference between HR and super-resolved video frames. To reduce complexity, model pruning can be used to minimize computational and storage requirements for model inference [6].…”

Section: Discussionmentioning

confidence: 99%

A New Dataset and Transformer for Stereoscopic Video Super-Resolution

Imani¹,

Islam²,

Wong³

2022

Preprint

View full text Add to dashboard Cite

Stereo video super-resolution (SVSR) aims to enhance the spatial resolution of the low-resolution video by reconstructing the high-resolution video. The key challenges in SVSR are preserving the stereo-consistency and temporalconsistency, without which viewers may experience 3D fatigue. There are several notable works on stereoscopic image super-resolution, but there is little research on stereo video super-resolution. In this paper, we propose a novel Transformer-based model for SVSR, namely Trans-SVSR. Trans-SVSR comprises two key novel components: a spatiotemporal convolutional self-attention layer and an optical flow-based feed-forward layer that discovers the correlation across different video frames and aligns the features. The parallax attention mechanism (PAM) that uses the crossview information to consider the significant disparities is used to fuse the stereo views. Due to the lack of a benchmark dataset suitable for the SVSR task, we collected a new stereoscopic video dataset, SVSR-Set, containing 71 full high-definition (HD) stereo videos captured using a professional stereo camera. Extensive experiments on the collected dataset, along with two other datasets, demonstrate that the Trans-SVSR can achieve competitive performance compared to the state-of-the-art methods. Project code and additional results are available at https://github. com/H-deep/Trans-SVSR/.

show abstract

Section: Discussionmentioning

confidence: 99%

A New Dataset and Transformer for Stereoscopic Video Super-Resolution

Imani¹,

Islam²,

Wong³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…The LASAO detector takes an average of 0.142 seconds to detect whether the subclavian artery is obstructed in a chest MRI image. Deep learning has enabled lots of artificial intelligence applications lately in image classification [4,7,13,15,19,22,24,27,33,[40][41][42][43]45] In this paper, last experiment is to probe the performances of the deep learning method to detect the left subclavian artery and aortic arch from a chest MRI image. It is difficult for a convolutional neural network model to precisely classify the objects in an image with multiple scales as well as small sizes, and complicated image background [19,43].…”

Section: Methodsmentioning

confidence: 99%

Identifying the occlusion of left subclavian artery with stent based on chest MRI images

Chan

Lin

Wang

et al. 2022

Multimed Tools Appl

View full text Add to dashboard Cite

Left subclavian artery occlusion is frequently treated by stent placement. The most commented on In-Stent Restenosis is excessive thrombus that squeezes the stent and leads to vessel occlusion. In this paper, an automatic system is provided to segment the left subclavian arteries and left aortic arches from chest MRI (Magnetic Resonance Imaging) images, and then to identify the occlusion of left subclavian artery based on the gray-levels of the extracted left subclavian artery and left aortic arch. Experimental results show that the system obtains the accuracy rate 97.33% of detecting the occlusion of left subclavian artery. The other task in this paper is to explore the relationship between stent/vascular diameter ratio and restenosis of left subclavian artery after stenting. Besides that, the experimental results also show that Assurance is useful for reducing the restenosis of left subclavian artery with stent.

show abstract

“…Several layers such as conv-layer and fully-connected layer have parameters whereas pooling and ReLU may not have parameters. The performance of a structured model after deployment is mostly dominated by the model complexity [10]. Lighter the model, faster the inferences and vice versa, generating a trade-off between the efficacy and performance matrix of a deployed structured model on tensor processing unit (TPU).…”

Section: Convolutional Neural Network and Pretrained Modelsmentioning

confidence: 99%

Object detection using YOLO: challenges, architectural successors, datasets and applications

Diwan

Anirudh

Tembhurne

2022

Multimed Tools Appl

427

View full text Add to dashboard Cite

Object detection is one of the predominant and challenging problems in computer vision. Over the decade, with the expeditious evolution of deep learning, researchers have extensively experimented and contributed in the performance enhancement of object detection and related tasks such as object classification, localization, and segmentation using underlying deep models. Broadly, object detectors are classified into two categories viz. two stage and single stage object detectors. Two stage detectors mainly focus on selective region proposals strategy via complex architecture; however, single stage detectors focus on all the spatial region proposals for the possible detection of objects via relatively simpler architecture in one shot. Performance of any object detector is evaluated through detection accuracy and inference time. Generally, the detection accuracy of two stage detectors outperforms single stage object detectors. However, the inference time of single stage detectors is better compared to its counterparts. Moreover, with the advent of YOLO (You Only Look Once) and its architectural successors, the detection accuracy is improving significantly and sometime it is better than two stage detectors. YOLOs are adopted in various applications majorly due to their faster inferences rather than considering detection accuracy. As an example, detection accuracies are 63.4 and 70 for YOLO and Fast-RCNN respectively, however, inference time is around 300 times faster in case of YOLO. In this paper, we present a comprehensive review of single stage object detectors specially YOLOs, regression formulation, their architecture advancements, and performance statistics. Moreover, we summarize the comparative illustration between two stage and single stage object detectors, among different versions of YOLOs, applications based on two stage detectors, and different versions of YOLOs along with the future research directions.

show abstract

Structured Model Pruning of Convolutional Networks on Tensor Processing Units

Cited by 11 publications

References 13 publications

A New Dataset and Transformer for Stereoscopic Video Super-Resolution

A New Dataset and Transformer for Stereoscopic Video Super-Resolution

Identifying the occlusion of left subclavian artery with stent based on chest MRI images

Object detection using YOLO: challenges, architectural successors, datasets and applications

Contact Info

Product

Resources

About