What is an object?

By Bogdan Alexe, Thomas Deselaers, Vittorio Ferrari. 2010

966 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

Scalable Object Detection Using Deep Neural Networks (2013, background): “ize. Each normalized coordinate is produced by a linear transformation of the last hidden layer. Conﬁdence: the conﬁdence score for the box containing an object is encoded as a single node value c i 2[0;1]. This value is produced through a linear transformation of the last hidden layer followed by a”
Training Region-Based Object Detectors with Online Hard Example Mining (2016, methodology): “and ImageNet detection. OverFeat is based on the sliding-window detection method, which is perhaps the most intuitive and oldest search method for detection. R-CNN, in contrast, uses region proposals [1,2,3,4,6,10,18,32,34], a method that was made popular by the selective search algorithm [32]. Sinc”
What Makes for Effective Detection Proposals? (2015, background): “Likewise, object proposals considerably reduce computation compared to the dense (sliding window) detection framework by generating candidate proposals that may contain objects.”
COCO-Stuff: Thing and Stuff Classes in Context (2016, methodology): “ing features: hue, saturation, value, objectness, and normalized image coordinates. To compute objectness for a pixel, we compute the objectness score for 10,000 bounding box proposals per image with [3]. We then deﬁne the objectness score for a pixel to be the average score over all boxes that cont”
Segment Anything (2023, background): “Segmentation is a broad ﬁeld: there’s interactive segmentation [57, 109], edge detection [3], super pixelization [85], object proposal generation [2], foreground segmentation [94], semantic segmentation [90], instance segmentation [66], panoptic segmentation [59], etc .”
Recurrent Models of Visual Attention (2014, background): “.g. [7, 24]), removing image regions from consideration via a branch and bound approach on the classiﬁer output (e.g. [13]), or by proposing candidate windows that are likely to contain objects (e.g. [1, 23]). Even though substantial speedups may be obtained with such approaches, and some of these c”
The Open Images Dataset V4 (2018, methodology): “To reduce the search space, the concept of “object proposals” (Alexe et al. 2010, 2012; Uijlings et al. 2013) was then introduced, which enable to work on just a few thousand windows instead of a dense grid.”
Scalable, High-Quality Object Detection (2014, background): “Given the recent success of novel sophisticated proposal generation methods [1, 22, 24, 14, 3, 10, 2, 11], the possibility of directly learned methods overtaking engineered meth-ods might seem far fetched at ﬁrst glance.”
On the Binding Problem in Artificial Neural Networks (2020, background): “Early approaches by-pass this problem by evaluating a fixed attention window exhaustively at each possible image location, or using several of many heuristics (Lampert et al., 2008; Alexe et al., 2010; Uijlings et al., 2013).”
Category-specific object reconstruction from a single image (2014, background): “The realization that convolutional networks [31] trained on huge datasets worked [29], together with the development of methods for generating object segmentation proposals [12, 1, 15, 43, 3, 28] has resulted in much sudden progress in object detection and segmentation [20, 23, 11].”
HyperNet: Towards Accurate Region Proposal Generation and Joint Object Detection (2016, background): “high recall with strict thresholds [28]. HyperNet achieves good results across a variety of IoU thresholds, which is desirable in practice and plays an important role in object detectors’ performance [1]. Figure 5 shows recall versus number of proposals for difRecall SelectiveSearch Edgeboxes RPN Hy”
R-CNN minus R (2015, methodology): “These methods, which are extensively reviewed in [13], originate in the idea of “objectness” proposed by [1].”
Watch and learn: Semi-supervised learning of object detectors from videos (2015, methodology): “ reliable short term tracking. 4.2. Reliable Tracking We formulate a scalable tracking procedure that effectively capitalizes on priors available from detection, color/texture consistency, objectness [1,51] and optical ﬂow. More importantly, our tracking procedure is very good at identifying its own”
Weakly Supervised Object Localization with Multi-Fold Multiple Instance Learning (2015, positive): “[1], [24], [49], [56], can be used to make MIL approaches to WSL for object localization manageable, and make it possible to use powerful and computationally expensive object models.”
Unidentified Video Objects: A Benchmark for Dense, Open-World Segmentation (2021, background): “The difference between objects and stuff has been discussed previously [18, 2, 4].”
How good are detection proposals, really? (2014, methodology): “The protocol introduced in [1] has served as a guideline for most other evaluations in the literature.”
Region-Based Saliency Detection and Its Application in Object Recognition (2014, methodology): “However, as we know, the image patches located in the object contribute more to the recognition of the object than the surrounding patches [2], [47].”
ME R-CNN: Multi-Expert Region-based CNN for Object Detection (2017, background): “[1] introduces a generic objectness measure, to estimate how likely it is for a region to contain object of any category using saliency, color contrast, edge density, and boundary information.”