Channel pruning has been long studied to compress convolutional neural networks (CNNs), which significantly reduces the overall computation. Prior works implement channel pruning in an unexplainable manner, which tends to reduce the final classification errors while failing to consider the internal influence of each channel. In this article, we conduct channel pruning in a white box. Through deep visualization of feature maps activated by different channels, we observe that different channels have a varying contribution to different categories in image classification. Inspired by this, we choose to preserve channels contributing to most categories. Specifically, to model the contribution of each channel to differentiating categories, we develop a class-wise mask for each channel, implemented in a dynamic training manner with respect to the input image's category. On the basis of the learned class-wise mask, we perform a global voting mechanism to remove channels with less category discrimination. Lastly, a fine-tuning process is conducted to recover the performance of the pruned model. To our best knowledge, it is the first time that CNN interpretability theory is considered to guide channel pruning. Extensive experiments on representative image classification tasks demonstrate the superiority of our White-Box over many state-of-the-arts (SOTAs). For instance, on CIFAR-10, it reduces 65.23% floating point operations per seconds (FLOPs) with even 0.62% accuracy improvement for ResNet-110. On ILSVRC-2012, White-Box achieves a 45.6% FLOP reduction with only a small loss of 0.83% in the top-1 accuracy for ResNet-50. Code is available at https://github.com/zyxxmu/White-Box.

Download full-text PDF

Source
http://dx.doi.org/10.1109/TNNLS.2022.3147269DOI Listing

Publication Analysis

Top Keywords

channel pruning
20
channel
8
pruning white
8
white box
8
image classification
8
class-wise mask
8
pruning
5
carrying cnn
4
cnn channel
4
box channel
4

Similar Publications

Optical Coherence Tomography (OCT) offers high-resolution images of the eye's fundus. This enables thorough analysis of retinal health by doctors, providing a solid basis for diagnosis and treatment. With the development of deep learning, deep learning-based methods are becoming more popular for fundus OCT image segmentation.

View Article and Find Full Text PDF

An Improved Neural Network Model Based on DenseNet for Fabric Texture Recognition.

Sensors (Basel)

December 2024

College of Science & Technology, Ningbo University, Ningbo 315300, China.

Article Synopsis
  • The accuracy of fabric texture identification is essential for automating knitted garment production and ensuring quality control, but traditional manual methods are inefficient and subjective.
  • To improve this, a new model called DLW-DenseNet is introduced, enhancing the DenseNet architecture with a learnable weight mechanism and differentiated learning strategy to optimize channel selection and reduce redundancy.
  • Additionally, a new dataset named KF9 for knitted fabrics has been created, and DLW-DenseNet has demonstrated a five-percentage-point increase in recognition accuracy compared to improved ResNet methods, proving its superiority in fabric texture recognition.
View Article and Find Full Text PDF

Lightweight object detection algorithms play a paramount role in unmanned aerial vehicles (UAVs) remote sensing. However, UAV remote sensing requires target detection algorithms to have higher inference speeds and greater accuracy in detection. At present, most lightweight object detection algorithms have achieved fast inference speed, but their detection precision is not satisfactory.

View Article and Find Full Text PDF

Detecting defects in complex urban sewer scenes is crucial for urban underground structure health monitoring. However, most image-based sewer defect detection models are complex, have high resource consumption, and fail to provide detailed damage information. To increase defect detection efficiency, visualize pipelines, and enable deployment on edge devices, this paper proposes a computer vision-based robotic defect detection framework for sewers.

View Article and Find Full Text PDF

Introduction: The accurate and rapid detection of ginseng fruits in natural environments is crucial for the development of intelligent harvesting equipment for ginseng fruits. Due to the complexity and density of the growth environment of ginseng fruits, some newer visual detection methods currently fail to meet the requirements for accurate and rapid detection of ginseng fruits. Therefore, this study proposes the YOLO-Ginseng detection method.

View Article and Find Full Text PDF

Want AI Summaries of new PubMed Abstracts delivered to your In-box?

Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!