ImageNet Auto-Annotation with Segmentation Propagation

By Matthieu Guillaumin, Daniel Küttel, Vittorio Ferrari. 2014

163 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

What Makes for Effective Detection Proposals? (2015, background): “For example, they can be used to handle unknown categories at test time, or to enable weakly supervised learning [57], [58], [59].”
Minding the Gaps for Block Frank-Wolfe Optimization of Structured SVMs (2016, methodology): “HorseSeg contains 25, 438 training images, 147 of which are manually annotated, 5, 974 annotations are constructed from object bounding boxes by the automatic method of Guillaumin et al. (2014), while the remaining 19, 317 annotations were constructed by the same method but without any human supervi”
BoxSup: Exploiting Bounding Boxes to Supervise Convolutional Networks for Semantic Segmentation (2015, methodology): “rom videos and improves detection accuracy. These experiments demonstrate the importance of dataset sizes for task-speciﬁc network training. For semantic segmentation, there have been existing papers [33,10] that investigate exploiting bounding box annotations instead of masks. But the box-level ann”
Weakly-and Semi-Supervised Learning of a Deep Convolutional Network for Semantic Image Segmentation (2015, methodology): “Bounding box annotations have been utilized for semantic segmentation by [38, 42], while [15, 21, 40] describe schemes exploiting both image-level labels and bounding box annotations.”
What's the Point: Semantic Segmentation with Point Supervision (2015, background): “ [26], free-form squiggles [17,18], noisy web tags [27], size constraints on objects [6] or heterogeneous annotations [33]. Common settings are image-level labels [4,23,25] and bounding boxes [4,34]. [14,35,36] use co-segmentation methods trained from image-level labels to automatically infer the se”
No More Discrimination: Cross City Adaptation of Road Scene Segmenters (2017, methodology): “Finally, [10] utilize image-level labels with cosegmentation techniques to infer semantic segmentation of foreground objects in the images of ImageNet.”
Salient Deconvolutional Networks (2016, methodology): “Combined with GrabCut, we achieve near state-of-the-art segmentation results on the ImageNet segmentation task of [4], while using off-the-shelf CNNs pretrained from a largely disjoint subset of ImageNet and with only image-level supervision.”
I Have Seen Enough: Transferring Parts Across Categories (2016, methodology): “Fewer works consider transfer learning for localization as we do; these include the works of [9, 19, 28, 33] that transfer bounding box information using the ImageNet hierarchies; the method of [22] that transfer object detectors from seed classes; and [1] which transfers detectors assuming a limite”
Object localization in ImageNet by looking out of the window (2015, methodology): “We evaluate our method on a subset of ImageNet containing 219 classes with more than 92000 images [15, 16, 35].”
Fine-Grained Image Analysis With Deep Learning: A Survey (2021, methodology): “Cosegmentation [120] based methods are also commonly used in this weakly supervised case.”
Training Object Class Detectors from Eye Tracking Data (2014, methodology): “As demonstrated in [18], the accuracy of the superpixel model is essentially identical to the corresponding pixel model.”
KITTI-360: A Novel Dataset and Benchmarks for Urban Scene Understanding in 2D and 3D (2021, methodology): “the per pixel annotation time of individual images, including classical methods [38], [60] and learning based methods [2], [3], [19], [59].”
Visualization of Supervised and Self-Supervised Neural Networks via Attribution Guided Factorization (2020, methodology): “We evaluate our method using three common datasets: (i) the validation set of ImageNet (Russakovsky et al. 2015) (ILSVRC) 2012, consisting of 50K images from 1000 classes, (ii) an annotated subset of ImageNet called ImageNet-Segmentation (Guillaumin, Küttel, and Ferrari 2014), containing 4,276 imag”
BigDatasetGAN: Synthesizing ImageNet with Pixel-wise Annotations (2022, background): “Reducing labeling costs can be achieved in a variety of ways, such as interactive humanin-the-loop annotation [2, 50, 53, 64, 73], nearest-neighbor mask transfer [26, 51, 63, 70], or using weak supervision in the form of cheaper labels such as object boxes [7, 43, 45], scribbles [48, 69] or very coa”
Transformer Interpretability Beyond Attention Visualization (2020, methodology): “Segmentation The segmentation metrics (pixel-accuracy, mAP, and mIoU) on ImageNet-segmentation are shown in Tab.”