Bottom-Up Segmentation for Top-Down Detection

By Sanja Fidler, Roozbeh Mottaghi, Alan L. Yuille, Raquel Urtasun. 2013

146 citations recorded by the source corpus.

Subfields: 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 Computer Vision and Pattern Recognition

Citations in the corpus

Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation (2013, methodology): “That is, we compute the unit’s activations on a large set of held-out region proposals (about 10 million), sort the proposals from highest to lowest activation, perform nonmaximum suppression, and then display the top-scoring regions.”
Hypercolumns for object segmentation and fine-grained localization (2014, methodology): “We concatenate features from some or all of the feature maps in the network into one long vector for every location which we call the hypercolumn at that location.”
Deformable part models are convolutional neural networks (2014, neutral): “SegDPM [13] 61.”
Simultaneous Detection and Segmentation (2014, methodology): “Since SegDPM detections are only available on PASCAL VOC2010 val, we restrict our evaluations only to this set.”
Detect What You Can: Detecting and Representing Objects Using Holistic Models and Body Parts (2014, methodology): “In the table, we show the results of our method using two different classifiers ([9] and [12]) for obtaining the unary appearance terms for the holistic object node.”
segDeepM: Exploiting segmentation and context in deep neural networks for object detection (2015, methodology): “Our work is inspired by the success of segDPM [8].”
Improving object detection with deep convolutional networks via Bayesian optimization and structured prediction (2015, methodology): “[16] incorporated the geometric consistency of bounding boxes with bottom-up segmentation as auxiliary features into the DPM.”
Beat the MTurkers: Automatic Image Labeling from Weak 3D Supervision (2014, background): “This is due to the availability of new sources of labeled data as well as the development of richer, holistic representations that combine multiple tasks [9, 15, 37].”
RIFD-CNN: Rotation-Invariant and Fisher Discriminative Convolutional Neural Networks for Object Detection (2016, methodology): “The DPM [30] and its variants [20, 22, 23, 34-39] have been the leading methods for object detection tasks for years owing to the carefully crafted features like HOG [32].”
Regionlets for Generic Object Detection (2013, unknown): “Performing object detection with object proposals has been attracting increasing attentions [ 31], [32], [33], [34], [35], [8].”
Towards Unified Object Detection and Semantic Segmentation (2014, unknown): “By concatenating the parameters for all mixtures as in [19,20], the score of a configuration can be written as”
Beyond Bounding Boxes: Precise Localization of Objects in Images (2015, background): “To see if this is the case, we compare to the SegDPM work of Fidler et al.[22].”
RIGOR: Reusing Inference in Graph Cuts for Generating Object Regions (2014, methodology): “It has also been used to generate additional features for bounding box detectors [11] and in video segmentation [24].”
BlitzNet: A Real-Time Deep Network for Scene Understanding (2017, background): “The idea of joint semantic segmentation and object detection was investigated ﬁrst for shallow approaches in [3, 18, 29, 7].”
The Role of Context for Object Detection and Semantic Segmentation in the Wild (2014, methodology): “Following the success of [13], we exploit both appearance and semantic segmentation as potentials in our model.”
Recent Advances in Deep Learning for Object Detection (2019, methodology): “Compared to the previous state-of-the-art methods based on a traditional detection framework SegDPM [44] with 40.”
Contextual Priming and Feedback for Faster R-CNN (2016, methodology): “[30] proposed to use segmentation proposals as additional features for DPM detection hypothesis.”
3D Object Proposals for Accurate Object Class Detection (2015, methodology): “The proposed solutions are then scored using simple Gestalt-like features, and typically only 150 top-ranked proposals are needed to succeed in consequent recognition tasks [21, 22, 7].”