Saved in:
Bibliographic Details
Main Authors: Mahmud, Tanvir, Yaman, Burhaneddin, Liu, Chun-Hao, Marculescu, Diana
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2403.16020
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917711156609024
author Mahmud, Tanvir
Yaman, Burhaneddin
Liu, Chun-Hao
Marculescu, Diana
author_facet Mahmud, Tanvir
Yaman, Burhaneddin
Liu, Chun-Hao
Marculescu, Diana
contents As deep neural networks evolve from convolutional neural networks (ConvNets) to advanced vision transformers (ViTs), there is an increased need to eliminate redundant data for faster processing without compromising accuracy. Previous methods are often architecture-specific or necessitate re-training, restricting their applicability with frequent model updates. To solve this, we first introduce a novel property of lightweight ConvNets: their ability to identify key discriminative patch regions in images, irrespective of model's final accuracy or size. We demonstrate that fully-connected layers are the primary bottleneck for ConvNets performance, and their suppression with simple weight recalibration markedly enhances discriminative patch localization performance. Using this insight, we introduce PaPr, a method for substantially pruning redundant patches with minimal accuracy loss using lightweight ConvNets across a variety of deep learning architectures, including ViTs, ConvNets, and hybrid transformers, without any re-training. Moreover, the simple early-stage one-step patch pruning with PaPr enhances existing patch reduction methods. Through extensive testing on diverse architectures, PaPr achieves significantly higher accuracy over state-of-the-art patch reduction methods with similar FLOP count reduction. More specifically, PaPr reduces about 70% of redundant patches in videos with less than 0.8% drop in accuracy, and up to 3.7x FLOPs reduction, which is a 15% more reduction with 2.5% higher accuracy. Code is released at https://github.com/tanvir-utexas/PaPr.
format Preprint
id arxiv_https___arxiv_org_abs_2403_16020
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle PaPr: Training-Free One-Step Patch Pruning with Lightweight ConvNets for Faster Inference
Mahmud, Tanvir
Yaman, Burhaneddin
Liu, Chun-Hao
Marculescu, Diana
Computer Vision and Pattern Recognition
As deep neural networks evolve from convolutional neural networks (ConvNets) to advanced vision transformers (ViTs), there is an increased need to eliminate redundant data for faster processing without compromising accuracy. Previous methods are often architecture-specific or necessitate re-training, restricting their applicability with frequent model updates. To solve this, we first introduce a novel property of lightweight ConvNets: their ability to identify key discriminative patch regions in images, irrespective of model's final accuracy or size. We demonstrate that fully-connected layers are the primary bottleneck for ConvNets performance, and their suppression with simple weight recalibration markedly enhances discriminative patch localization performance. Using this insight, we introduce PaPr, a method for substantially pruning redundant patches with minimal accuracy loss using lightweight ConvNets across a variety of deep learning architectures, including ViTs, ConvNets, and hybrid transformers, without any re-training. Moreover, the simple early-stage one-step patch pruning with PaPr enhances existing patch reduction methods. Through extensive testing on diverse architectures, PaPr achieves significantly higher accuracy over state-of-the-art patch reduction methods with similar FLOP count reduction. More specifically, PaPr reduces about 70% of redundant patches in videos with less than 0.8% drop in accuracy, and up to 3.7x FLOPs reduction, which is a 15% more reduction with 2.5% higher accuracy. Code is released at https://github.com/tanvir-utexas/PaPr.
title PaPr: Training-Free One-Step Patch Pruning with Lightweight ConvNets for Faster Inference
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2403.16020