Mimicking Very Efficient Network for Object Detection

By Quanquan Li, Sheng Jin, Junjie Yan. 2017

374 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Knowledge Distillation: A Survey (2020, background): “, 2020a), object detection (Li et al., 2017; Hong and Yu, 2019; Shmelkov et al., 2017; Wei et al., 2018; Wang et al., 2019d), lane detection (Hou et al.”
Deformable ConvNets V2: More Deformable, Better Results (2018, methodology): “Given a large teacher model, a compact student model is trained by mimicking the teacher model output or feature responses on training images [1, 21, 26].”
Data-Free Learning of Student Networks (2019, background): “[13] further presented a feature mimic framework to train efficient convolutional networks for objective detection.”
Decoupled Knowledge Distillation (2022, neutral): “This rule also stands in transferring knowledge between detectors [17, 37], i.e ., feature mimicking is of vital importance since logits are not capable of providing knowledge for object localization.”
A Survey of Deep Learning-Based Object Detection (2019, methodology): “Some methods [24], [87], [121]–[124] are aim to reach real-time detection.”
Object Detection in 20 Years: A Survey (2019, methodology): “Recently, this idea has been used in the acceleration of object detection [195, 196].”
Knowledge Distillation via Route Constrained Optimization (2019, background): “In several works [26, 17] the KL divergence is replaced by euclidean distance,”
Structured Knowledge Distillation for Semantic Segmentation (2019, methodology): “• Feature distillation by MIMIC [35, 21]: We follow [21] to align the features of each pixel between T and S through a 1 × 1 convolution layer to match the dimension of the feature”
Distilling Object Detectors With Fine-Grained Feature Imitation (2019, background): “[20] proposed to only transfer knowledge under the area of proposals, but the mimicking regions depend on the output of model itself and it is not applicable for onestage detector.”
Channel-wise Knowledge Distillation for Dense Prediction* (2020, methodology): “Previous research [24, 19] found that directly transferring the KD methods [15, 2] in classification to semantic segmentation may not lead to satisfactory results.”
Knowledge Distillation for End-to-End Person Search (2019, neutral): “When moving from classification to detection, as only few works [8, 30, 46] attempted, complex modelling questions arise as to where and how to supervise, which are not entirely answered yet.”
Focal and Global Knowledge Distillation for Detectors (2021, methodology): “[15] choose the features sampled from RPN to calculate distillation loss.”
Recent Advances in Deep Learning for Object Detection (2019, background): “In object detection, some works [164, 132] also investigate this training scheme to improve detection performance.”
Deep Face Recognition: A Survey (2018, background): “Moreover, some well-known compressed networks such as Pruning [264], [265], [266], BinaryNets [267], [268], [269], [270], Mimic Networks [271], [272], also have potential to be introduced into FR.”
Object Detection With Deep Learning: A Review (2018, background): “And this situation can be relieved by introducing better pre-training schemes [225], knowledge distillation [226] and hint learning [227].”
Convolutional neural network: a review of models, methodologies and applications to object detection (2019, background): “VGGNet 2017 Quanquan Li [97] Highest precision is 97.”
Diverse Knowledge Distillation for End-to-End Person Search (2020, methodology): “It is a common practice to improve the main network using the additional guidance from external networks (Li, Jin, and Yan 2017; Cheng et al. 2018; Zhu et al. 2019; Hinton, Vinyals, and Dean 2015; Tung and Mori 2019; Liu et al. 2019; Yuan et al. 2020).”
Localization Distillation for Object Detection (2022, background): “[36] proposed to mimic the features within the region proposal for Faster R-CNN.”