Convolutional Channel Features

By Binh Yang, Junjie Yan, Zhen Lei, S. Li. IEEE International Conference on Computer Vision, 2015

308 citations recorded by the source corpus.

Deep learning methods are powerful tools but often suffer from expensive computation and limited flexibility. An alternative is to combine light-weight models with deep representations. As successful cases exist in several visual problems, a unified framework is absent. In this paper, we revisit two widely used approaches in computer vision, namely filtered channel features and Convolutional Neural Networks (CNN), and absorb merits from both by proposing an integrated method called Convolutional Channel Features (CCF). CCF transfers low-level features from pre-trained CNN models to feed the boosting forest model. With the combination of CNN features and boosting forest, CCF benefits from the richer capacity in feature representation compared with channel features, as well as lower cost in computation and storage compared with end-to-end CNN methods. We show that CCF serves as a good way of tailoring pre-trained CNN models to diverse tasks without fine-tuning the whole network to each task by achieving state-of-the-art performances in pedestrian detection, face detection, edge detection and object proposal generation.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/609b60fb5cf6c45d47173a49e8e55792c046ee80

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

Citations in the corpus

WIDER FACE: A Face Detection Benchmark (2015, unknown): “Modern face detection algorithms can be categorized into four categories: cascade based methods [2, 10, 15, 16, 21], part based methods [19, 23, 30], channel feature based methods [25, 24], and neural network based methods [6, 14, 25, 28].”
Semantic-Aware Scene Recognition (2019, background): “Convolutional Channel Features (CCF) [36] uses extracted low-level features from VGG-16 network as an additional input channel to enhance pedestrian detection.”
A comprehensive review on ensemble deep learning: Opportunities and challenges (2023, methodology): “In Yang et al. (2015) proposed boosting CNNs for face detection using imageNet dataset (Krizhevsky et al., 2012).”
What Can Help Pedestrian Detection? (2017, methodology): “CCF [32] uses the activation maps of a VGG-16 [25] network pretrained on ImageNet [16] as channel feature, while Costea and Nedevschi [8] utilize the heatmap of semantic scene parsing, in which detectors benefit from the semantic information within a large receptive field.”
Repulsion Loss: Detecting Pedestrians in a Crowd (2017, methodology): “Recently, the CNN-based detectors [16, 28, 19, 13, 26] show great potential in dominating the ﬁeld of pedestrian detection.”
Towards Robust Deep Neural Networks with BANG (2016, background): “Machine learning is broadly used in various real-world vision applications and recent advances in deep learning have made deep neural networks the most powerful learning models that can be successfully applied to different vision problems [27, 25, 7, 28, 20, 14, 18, 17, 29].”
Knowledge Distillation for End-to-End Person Search (2019, unknown): “There is a large body of work on person detection, from methods based on hand-crafted features [16, 18] to deep convolutional neural network (CNN) feature-learning methods [19, 20, 60].”
SSH: Single Stage Headless Face Detector (2017, background): “Face detectors based on CNNs significantly closed the performance gap between human and artificial detectors [12, 33, 32, 38, 7].”
Person Re-identification in the Wild (2016, background): “Oriented Gradient (HOG) [9] as well as the recently proposed Convolutional Channel Features (CCF) [44].”
End-to-End Deep Learning for Person Search (2016, methodology): “ commonly used off-the-shelf pedestrian detectors. They rely on hand-crafted features and linear classiﬁers to detect pedestrians. Recent years, CNNbased pedestrian detectors have also been developed [37, 40]. Various factors, including CNN model structures, training data, and different training str”
Handcrafted vs. non-handcrafted features for computer vision classification (2017, methodology): “Paper Features extracted using CNN here Using deep layers, shallow layers, and the scores obtained by CNN [14] Representation of images as strings of the top layers of pretrained CNNs [15] The convolutional layers of a CNN are used as a filter bank [16] Representation of images using the seventh ful”
Object Detection With Deep Learning: A Review (2018, unknown): “Joint Cascade [30], SURF-multiview [184], Viola-Jones [166], NPDFace [185], Faceness [169], CCF [186], MTCNN [176], Conv3D [175], Hyperface [187], UnitBox [167], LDCF+ [S2], DeepIR [173], HR-ER [188], Face-R-CNN [172] and ScaleFace [170].”
HyperFace: A Deep Multi-Task Learning Framework for Face Detection, Landmark Localization, Pose Estimation, and Gender Recognition (2019, methodology): “Some of the recent published methods compared in our evaluations include DP2MFD [42], Faceness [60], HeadHunter [39], JointCascade [6], CCF [59], SquaresChnFtrs-”
Detecting Masked Faces in the Wild with LLE-CNNs (2017, methodology): “[29] proposed a face detector that utilized the feature aggregation framework [3], while the features were generated via CNN.”
Occlusion-aware R-CNN: Detecting Pedestrians in a Crowd (2018, methodology): “Recently, pedestrian detection is dominated by the CNN-based methods ( e.g. , [44,22,50,5,60,4]).”