ContextLocNet: Context-Aware Deep Network Models for Weakly Supervised Localization

By Vadim Kantorov, Maxime Oquab, Minsu Cho, Ivan Laptev. European Conference on Computer Vision, 2016

314 citations recorded by the source corpus.

We aim to localize objects in images using image-level supervision only. Previous approaches to this problem mainly focus on discriminative object regions and often fail to locate precise object boundaries. We address this problem by introducing two types of context-aware guidance models, additive and contrastive models, that leverage their surrounding context regions to improve localization. The additive model encourages the predicted object region to be supported by its surrounding context region. The contrastive model encourages the predicted object region to be outstanding from its surrounding context region. Our approach benefits from the recent success of convolutional neural networks for object recognition and extends Fast R-CNN to weakly supervised object localization. Extensive experimental evaluation on the PASCAL VOC 2007 and 2012 benchmarks shows that our context-aware approach significantly improves weakly supervised localization and detection.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/5cd8b2dcb25efd0cc6356230ee09a0e4d4caea5e

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Weakly Supervised Object Localization and Detection: A Survey (2021, unknown): “The whole weakly supervised learning framework may be designed in a compact network model, such as in [47], [77], [103],”
Hide-and-Seek: Forcing a Network to be Meticulous for Weakly-Supervised Object and Action Localization (2017, background): “Other network architectures have been designed for weakly-supervised object detection [20, 4, 24].”
Cross-Domain Weakly-Supervised Object Detection Through Progressive Domain Adaptation (2018, methodology): “For Faster R-CNN [28], we used the reimplementation provided in ChainerCV [25].”
Revisiting Knowledge Transfer for Training Object Class Detectors (2017, background): “ training images. This is tedious and very timeconsuming. Therefore, several research efforts have been devoted to training object detectors under weak supervision, i.e. using only image-level labels [2,4,7,8,21,29,36, 40,41,42,48]. While this is substantially cheaper, the resulting detectors typica”
Efficient Deep CNN-Based Fire Detection and Localization in Video Surveillance Applications (2019, methodology): “CNNs have shown encouraging performance in numerous 215 computer vision problems and applications, such as object 216 detection and localization [28], [29], image segmentation, 217 super-resolution, classification [30]–[33], and indexing and 218 retrieval [34].”
Attention-Based Dropout Layer for Weakly Supervised Object Localization (2019, background): “Existing approaches mine and track discriminative features of each class for object detection [45, 36, 37, 9, 45, 25, 21, 41, 19, 2, 39, 15, 63, 7, 5, 4, 48, 14, 65, 32, 31, 58, 62, 8, 6] and segmentation [33, 29, 18, 16, 24, 52, 50].”
Min-Entropy Latent Model for Weakly Supervised Object Detection (2018, methodology): “[27], Self-Taught [29], WCCN [28], WeakRPN [60], and TS [25] train/val 23.8 WCCN [28] trainval/test 28.4 ContextNet [24] trainval/test 35.3 OICR-VGGM [27] trainval VGG16 PDA [23] train/val 29.1 Self-Taught [29] train/val 39.0 WCCN [28] trainval/test 37.9 OICR [27] trainval/test 37.9 Self-Taught…”
Y Qualification Main task Y Return N N Y User qualified ? N Feedback Pass ? Pass ? Annotator training Annotating images (2017, methodology): “State-of-the-art WSOL approaches based on AlexNet architecture [34] perform as follows.”
LayerCAM: Exploring Hierarchical Class Activation Maps for Localization (2021, background): “Some researchers [47], [48], [49], [50], [51], [52] attempt to solve WSOL as a multiple instance learning framework.”
C-MIL: Continuation Multiple Instance Learning for Weakly Supervised Object Detection (2019, methodology): “To alleviate this problem, researchers introduced spatial regularization [8,14,37], context information [22, 38], and progressive optimization [14, 15, 34, 37, 40] into the MIL networks.”
Collaborative Learning for Weakly Supervised Object Detection (2018, background): “, 2014] and various end-to-end MIL-CNN models [Bilen and Vedaldi, 2016; Kantorov et al., 2016; Tang et al., 2017] following the twostream structure of WSDDN [Bilen and Vedaldi, 2016].”
CAD-Net: A Context-Aware Detection Network for Objects in Remote Sensing Imagery (2019, background): “For object detection, several studies [35]–[37] have shown that contextual information can help improve the object detection performance, especially in the presence of small objects.”
Early fire detection using convolutional neural networks during surveillance for effective disaster management (2017, background): “Convolutional neural networks have exhibited state-of-the-art performance in a variety of computer vision tasks including image classiﬁcation and retrieval [28–30] , object detection [31,32] , localization [33] , and image segmentation [34] .”
Global Weighted Average Pooling Bridges Pixel-level Localization and Image-level Classification (2018, methodology): “Based on this framework, context-aware network [22], expectationmaximization algorithm [23], online instance classifier refinement process [24] and collaborative learning method [25] are proposed to improve the performance.”
A Survey of Deep Learning-Based Object Detection (2019, methodology): “Context information [209], instance classifier refinement [210] and image segmentation [211] [212] are adopted to tackle hardly optimized problems.”
Learned Contextual Feature Reweighting for Image Geo-Localization (2017, background): “Contextual information has been widely used for object recognition [39, 42, 64, 8, 13, 27].”
From Discriminant to Complete: Reinforcement Searching-Agent Learning for Weakly Supervised Object Detection (2020, methodology): “The compared state-of-the-art methods include both the methods using the pretrained CNN features followed by the carefully-designed shallow learning models [4], [6], [7], and the methods directly learn the WSODs in the end-to-end manner [5], [9], [51].”