DSFD: Dual Shot Face Detector

By Jian Li, Yabiao Wang, Changan Wang, Ying Tai, Jianjun Qian, Jian Yang, Chengjie Wang, Jilin Li, Feiyue Huang. 2018

481 citations recorded by the source corpus.

Recently, Convolutional Neural Network (CNN) has achieved great success in face detection. However, it remains a challenging problem for the current face detection methods owing to high degree of variability in scale, pose, occlusion, expression, appearance and illumination. In this Paper, we propose a novel detection network named Dual Shot face Detector(DSFD). which inherits the architecture of SSD and introduces a Feature Enhance Module (FEM) for transferring the original feature maps to extend the single shot detector to dual shot detector. Specially, progressive anchor loss (PAL) computed by using two set of anchors is adopted to effectively facilitate the features. Additionally, we propose an improved anchor matching (IAM) method by integrating novel data augmentation techniques and anchor design strategy in our DSFD to provide better initialization for the regressor. Extensive experiments on popular benchmarks: WIDER FACE (easy: 0.966, medium: 0.957, hard: 0.904) and FDDB ( discontinuous: 0.991, continuous: 0.862 ) demonstrate the superiority of DSFD over the state-of-the-art face detection methods (e.g., PyramidBox and SRN). Code will be made available upon publication.

Subfields: Computer Vision and Pattern Recognition, Signal Processing

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

Visual Programming: Compositional visual reasoning without training (2022, methodology): “Red modules use neural models (OWL-ViT [21], DSFD [18], MaskFormer [6], CLIP [23], ViLT [16], and Stable Diffusion [28]).”
Advancing Image Understanding in Poor Visibility Environments: A Collective Benchmark Study (2019, unknown): “1) Baseline Composition: We test four state-of-the-art deep face detectors: (1) Dual Shot Face Detector ( DSFD ) [147] 10 ; (2) Pyramidbox [148] 11 ; (3) Single Stage Headless Face Detector ( SSH ) [149] 12 ; (4) Faster RCNN [150].”
Scale Match for Tiny Person Detection (2019, unknown): “Therefore, the state-of-the-art of DSFD detector [13], which is speciﬁed for tiny face detection, has been included for comparison on TinyPerson.”
Recent Advances in Deep Learning for Object Detection (2019, background): “After the progresses of deep learning in image classification, face detectors based on deep learning significantly outperformed traditional face detectors [183, 184, 185, 186, 187].”
YOLO-FaceV2: A Scale and Occlusion Aware Face Detector (2022, methodology): “Anchor can provide good priori information, thus using denser anchors and corresponding matching strategies can eﬀectively improve the quality of object proposals [24, 25, 37, 40].”
DeepPrivacy: A Generative Adversarial Network for Face Anonymization (2019, methodology): “ in cases where faces from "hard" and easy/medium are present in the same image. Experiment Details: For the face detector we use the current state-ofthe-art, Dual Shot Face Detector (DSFD) [14]. The WIDER-Face dataset has no facial keypoint annotations; therefore, we automatically detect ”
Learning to Enhance Low-Light Image via Zero-Reference Deep Curve Estimation (2021, methodology): “16, after image enhancement, the precision of DSFD [48] increases considerably compared to that using the input images without enhancement.”
Multi-task Generative Adversarial Network for Detecting Small Objects in the Wild (2020, methodology): “…handcrafted feature based methods (Viola and Jones 2004; Yan et al. 2014, 2013; Zhang et al. 2017c) and CNN-based methods (Cai et al. 2016; Jiang and Learned-Miller 2017; Wan et al. 2016; Zhu et al. 2017a; Bai and Ghanem 2017; Bai et al. 2018a; Wang et al. 2017a; Li et al. 2018; Wen et al. 2019).”
Condensed Movies: Story Based Retrieval with Contextual Embeddings (2020, methodology): “Face embeddings for each face track are extracted in three stages: (1) Each frame is passed through a dual shot face detector [54] (trained on the Wider Face dataset [55]) to extract bounding boxes.”
Benchmarking Low-Light Image Enhancement and Beyond (2021, methodology): “It is clearly demonstrated that, after ﬁnetuning or retraining on VE-LOL-L, the models improve the performance a lot, such as the proposed one, ﬁnetuned DSFD and the version without MDL.”
R2RNet: Low-light Image Enhancement via Real-low to Real-normal Network (2021, methodology): “Because there are no corresponding labels in the test set, we randomly select 100 images from the training set for evaluation, applying our R2RNet as a pre-processing step, followed by two state-of-art pre-trained face detection methods: RetinaFace [55] and DSFD [56].”
Zero-Reference Deep Curve Estimation for Low-Light Image Enhancement (2020, methodology): “A state-of-the-art deep face detector, Dual Shot Face Detector (DSFD) [14], trained on WIDER FACE dataset [28], is used as the baseline model.”
High-Level Semantic Feature Detection: A New Perspective for Pedestrian Detection (2019, background): “As stated in [26], the detection head plays a significant role in top performance, which has been extensively explored in the literature [10, 26, 20, 19].”
Towards Large-Scale Small Object Detection: Survey and Benchmarks (2022, methodology): “Following this roadmap, DSFD [71] employs two-shot detector connected by the feature enhancement module to detect the faces of various scales.”
RetinaFace: Single-stage Dense Face Localisation in the Wild (2019, unknown): “[71], Pyramid-Box [49], FDNet [63], SRN [8], FANet [65], DSFD [27], DFS [50], VIM-FD [69], ISRN [67]).”