Pedestrian Detection with Unsupervised Multi-stage Feature Learning

By Pierre Sermanet, Koray Kavukcuoglu, Soumith Chintala, Yann LeCun. 2012

834 citations recorded by the source corpus.

Pedestrian detection is a problem of considerable practical interest. Adding to the list of successful applications of deep learning methods to vision, we report state-of-the-art and competitive results on all major pedestrian datasets with a convolutional network model. The model uses a few new twists, such as multi-stage features, connections that skip layers to integrate global shape information with local distinctive motif information, and an unsupervised method based on convolutional sparse coding to pre-train the filters at each stage.

Subfields: Biomedical Engineering, 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 Biomedical Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Fully convolutional networks for semantic segmentation (2014, background): “e of ﬁxed smoothing and differencing. The jet of Koenderink and van Doorn [27] is a rich, local feature deﬁned by compositions of partial derivatives. In the context of deep networks, Sermanet et al. [28] fuse intermediate layers but discard resolution in doing so. In contemporary work Hariharan et ”
OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks (2013, background): “More recently, ConvNets have been shown to yield state of the art performance on text detection in natural images [4], face detection [8, 23] and p e estrian detection [25].”
Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation (2013, methodology): “To adapt our CNN to the new task (detection) and the new domain (warped proposal windows), we continue stochastic gradient descent (SGD) training of the CNN parameters using only warped region proposals.”
Playing Atari with Deep Reinforcement Learning (2013, background): “Recent advances in deep learning have made it possible to extract high-level features from raw sensory data, leading to breakthroughs in computer vision [11, 22, 16] and speech recognition [6, 7].”
Benchmarking Detection Transfer Learning with Vision Transformers (2021, background): “Notably, object detection, which has played a central role in the study of transfer learning over the last decade ( e.g ., [35, 14, 9, 17]), was not explored in the pioneering work on ViT training [10, 7, 5]—supervised or unsupervised—due to the challenges (described shortly) of integrating ViTs…”
Hypercolumns for object segmentation and fine-grained localization (2014, background): “[34] combine subsampled intermediate layers with the top layer for pedestrian detection.”
Beyond Skip Connections: Top-Down Modulation for Object Detection (2016, methodology): “For example, [4, 10, 22, 42] combine features from multiple layers for the final classifier; while [4, 42] use subsampled features from finer scales, [10, 22] upsample the features to the finest scale and use their combination.”
Learning to Refine Object Segments (2016, methodology): “For pixel-labeling tasks such as semantic segmentation and edge detection, ‘skip’ connections [24, 29–31], as shown in Figure 1b, are popular.”
Inside-Outside Net: Detecting Objects in Context with Skip Pooling and Recurrent Neural Networks (2015, methodology): “These skip-layers [34, 23, 24, 12] span multiple spatial resolutions and levels of feature abstraction.”
Simultaneous Detection and Segmentation (2014, methodology): “They have since been used in a variety of tasks, including detection [29,28] and semantic segmentation [12].”
Learning to See by Moving (2015, background): “2) and (4) Object recognition on Imagenet [36] (section 5.”
A MultiPath Network for Object Detection (2016, methodology): “[25] used two contextual regions centered on each object for pedestrian detection.”
Unsupervised Learning of Visual Representations Using Videos (2015, methodology): “The third class of algorithms and more related to our paper is unsupervised learning of visual representations from the pixels themselves using deep learning approaches [21, 26, 44, 39, 29, 47, 9, 33, 2, 49, 8].”
Unsupervised Learning of Long-Term Motion Dynamics for Videos (2017, methodology): “In the RGB domain, unsupervised learning of visual representations has shown usefulness for various supervised tasks such as pedestrian detection and object detection [1, 26].”
Cross Modal Distillation for Supervision Transfer (2015, background): “Recent work as presented in [18, 31, 36] operationalize this and demonstrate improvements for fine grained tasks like object and part segmentation.”
Deep Networks with Internal Selective Attention through Feedback Connections (2014, background): “Deep convolutional neural networks (CNNs) [1] with max-pooling layers [2] trained by backprop [3] on GPUs [4] have become the state-of-the-art in object recognition [5, 6, 7, 8], segmentation/detection [9, 10], and scene parsing [11, 12] (for an extensive review see [13]).”
Deep learning for detecting robotic grasps (2013, methodology): “[57] applied a convolutional neural network for pedestrian detection, and Coates et al.”
PANDA: Pose Aligned Networks for Deep Attribute Modeling (2013, methodology): “has become recently available, convolutional nets have become the most accurate method for generic object category classification [15] and pedestrian detection [25].”