A Facial Landmark Detection Method Based on Deep Knowledge Transfer

Gao, Pengcheng; Lü, Ke; Xue, Jian; Lyu, Jiayi; Shao, Ling

doi:10.1109/tnnls.2021.3105247

Cited by 7 publications

(3 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Guo et al [ 27 ] trained a lightweight network consisting of the MobileNetV2 [ 36 ] blocks by using an auxiliary 3D pose estimator. To utilize the learning ability of large models, some recent works [ 28 , 29 , 30 , 31 ] used the teacher-guided KD technique to make a small student network learn the dark knowledge from a large teacher network. The student networks were usually based on the existing lightweight networks (e.g., MobileNetV2, EfficientNet-B0 [ 37 ], and HRNetV2-W9 [ 34 ]), while the teacher networks use the large CNN models (e.g., ResNet-50, EfficientNet-B7 [ 37 ], and HRNetV2-W18 [ 34 ]) as the network backbone.…”

Section: Related Workmentioning

confidence: 99%

“…Recently, some researchers have tended to balance the accuracy and efficiency of a facial landmark detector. They either train a small model from scratch [ 26 , 27 ] or use knowledge distillation (KD) for model compression [ 28 , 29 , 30 , 31 ]. The former aims to design a lightweight network combined with an effective learning strategy, while the latter considers how to apply the KD technique to transfer the dark knowledge from a large network to a small one.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

SD-HRNet: Slimming and Distilling High-Resolution Network for Efficient Face Alignment

Lin

Zheng

Zhao

et al. 2023

Sensors

View full text Add to dashboard Cite

Face alignment is widely used in high-level face analysis applications, such as human activity recognition and human–computer interaction. However, most existing models involve a large number of parameters and are computationally inefficient in practical applications. In this paper, we aim to build a lightweight facial landmark detector by proposing a network-level architecture-slimming method. Concretely, we introduce a selective feature fusion mechanism to quantify and prune redundant transformation and aggregation operations in a high-resolution supernetwork. Moreover, we develop a triple knowledge distillation scheme to further refine a slimmed network, where two peer student networks could learn the implicit landmark distributions from each other while absorbing the knowledge from a teacher network. Extensive experiments on challenging benchmarks, including 300W, COFW, and WFLW, demonstrate that our approach achieves competitive performance with a better trade-off between the number of parameters (0.98 M–1.32 M) and the number of floating-point operations (0.59 G–0.6 G) when compared to recent state-of-the-art methods.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

SD-HRNet: Slimming and Distilling High-Resolution Network for Efficient Face Alignment

Lin

Zheng

Zhao

et al. 2023

Sensors

View full text Add to dashboard Cite

show abstract

“…Hannane et al [39] learned a FLM topological model that performs divide-conquer search for different patches of the face using coarse to fine CNN techniques and subsequently refines the landmarks positions by using a shallow cascaded CNN regression. Gao has developed a supervised encoder-decoder architecture [40] based on EfficientNet-B0 where the dark knowledge extracted from teacher network is used to supervise the training of a small student network and patch similarity (PS) distillation is used learn the structural information of the face.…”

Section: Coarse-to-fine Techniquesmentioning

confidence: 99%