Focal and Global Knowledge Distillation for Detectors

Yang, Zhendong; Li, Zhe; Jiang, Xiaohu; Gong, Yuan; Yuan, Zehuan; Zhao, Danpei; Yuan, Chun

doi:10.1109/cvpr52688.2022.00460

Cited by 187 publications

(61 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Due to the large-scale variation of objects in industrial detection datasets and the uneven distribution of positive and negative samples of objects, these will adversely affect the distillation effect. For this reason, this study uses the scaling mask to balance the object scale with reference to FGD [ 44 ] to solve the problem of an unbalanced object scale. The formula is as follows:

where GT is the region of the ground truth.…”

Section: Methodsmentioning

confidence: 99%

A Defect Detection Model for Industrial Products Based on Attention and Knowledge Distillation

Zhang

Zhou

Shao

et al. 2022

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

Industrial quality detection is one of the important fields in machine vision. Big data analysis, the Internet of Things, edge computing, and other technologies are widely used in industrial quality detection. Studying an industrial detection algorithm that can be organically combined with the Internet of Things and edge computing is imminent. Deep learning methods in industrial quality detection have been widely proposed recently. However, due to the particularity of industrial scenarios, the existing deep learning-based general object detection methods have shortcomings in industrial applications. This study designs two isomorphic industrial detection models to solve these problems: T-model and S-model. Both proposed models combine swin-transformer with convolution in the backbone and design a residual fusion path. In the neck, this study designs a dual attention module to improve feature fusion. Second, this study presents a knowledge distiller based on the dual attention module to improve the detection accuracy of the lightweight S-model. According to the analysis of the experimental results on four public industrial defect detection datasets, the model in this study is more advantageous in industrial defect detection.

show abstract

where GT is the region of the ground truth.…”

Section: Methodsmentioning

confidence: 99%

A Defect Detection Model for Industrial Products Based on Attention and Knowledge Distillation

Zhang

Zhou

Shao

et al. 2022

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

show abstract

“…Dai et al [26] developed GID framework which selects distillation areas based on differences between the student and teacher networks. Yang et al proposed FGD [5] which separates the foreground and background, enabling the student model to learn from the teacher network areas of interest and global knowledge via local and global distillation respectively. Besides, MGD [6] imposes random masking on the feature map of the student model, and then generates the feature map reconstructing from the teacher network.…”

Section: B Knowledge Distillationmentioning

confidence: 99%

“…FGFI [4] operates by distilling the features of high IoU between ground truth and anchors. FGD [5] was developed to separate distillation of foreground and background. Recent research suggests it is preferable for the student model to reconstruct and learn expressive features from the teacher model in the first place instead of following the teacher for generating competitive representations.…”

Section: Introductionmentioning

confidence: 99%

AMD: Adaptive Masked Distillation for Object

Yang¹,

Tang²,

Li³

et al. 2023

Preprint

View full text Add to dashboard Cite

As a general model compression paradigm, featurebased knowledge distillation allows the student model to learn expressive features from the teacher counterpart. In this paper, we mainly focus on designing an effective feature-distillation framework and propose a spatial-channel adaptive masked distillation (AMD) network for object detection. More specifically, in order to accurately reconstruct important feature regions, we first perform attention-guided feature masking on the feature map of the student network, such that we can identify the important features via spatially adaptive feature masking instead of random masking in the previous methods. In addition, we employ a simple and efficient module to allow the student network channel to be adaptive, improving its model capability in object perception and detection. In contrast to the previous methods, more crucial object-aware features can be reconstructed and learned from the proposed network, which is conducive to accurate object detection. The empirical experiments demonstrate the superiority of our method: with the help of our proposed distillation method, the student networks report 41.3%, 42.4%, and 42.7% mAP scores when RetinaNet, Cascade Mask-RCNN and RepPoints are respectively used as the teacher framework for object detection, which outperforms the previous state-of-the-art distillation methods including FGD and MGD.

show abstract

“…55,56 Furthermore, KD is used for improving object detectors by selecting different valuable areas (e.g., foreground) to distill. [57][58][59] Inspired by these directions, we extend it to HOI recognition in videos, allowing knowledge transfer between global and local contextual views of interactions via KD.…”

Section: Knowledge Distillationmentioning

confidence: 99%

Distillation of human–object interaction contexts for action recognition

Almushyti

2022

Computer Animation & Virtual

View full text Add to dashboard Cite

Modeling spatial-temporal relations is imperative for recognizing human actions, especially when a human is interacting with objects, while multiple objects appear around the human differently over time. Most existing action recognition models focus on learning overall visual cues of a scene but disregard a holistic view of human-object relationships and interactions, that is, how a human interacts with respect to short-term task for completion and long-term goal. We therefore argue to improve human action recognition by exploiting both the local and global contexts of human-object interactions (HOIs). In this paper, we propose the Global-Local Interaction Distillation Network (GLIDN), learning human and object interactions through space and time via knowledge distillation for holistic HOI understanding. GLIDN encodes humans and objects into graph nodes and learns local and global relations via graph attention network. The local context graphs learn the relation between humans and objects at a frame level by capturing their co-occurrence at a specific time step.The global relation graph is constructed based on the video-level of human and object interactions, identifying their long-term relations throughout a video sequence. We also investigate how knowledge from these graphs can be distilled to their counterparts for improving HOI recognition. Finally, we evaluate our model by conducting comprehensive experiments on two datasets including Charades and CAD-120. Our method outperforms the baselines and counterpart approaches.

show abstract

Focal and Global Knowledge Distillation for Detectors

Cited by 187 publications

References 22 publications

A Defect Detection Model for Industrial Products Based on Attention and Knowledge Distillation

A Defect Detection Model for Industrial Products Based on Attention and Knowledge Distillation

AMD: Adaptive Masked Distillation for Object

Distillation of human–object interaction contexts for action recognition

Contact Info

Product

Resources

About