Multiple instance learning for soft bags via top instances
By Wei-Xin Li, Nuno Vasconcelos. 2015
87 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
- Luc Van Gool
- Dacheng Tao
- Xiaogang Wang
- Ming-Hsuan Yang
- Trevor Darrell
- Bernt Schiele
- Andrew Zisserman
- Andrea Vedaldi
Citations in the corpus
- Video Representation Learning Using Discriminative Pooling (2018, methodology): “An important component of our scheme is the MIL scheme, which is a popular data selection technique [7, 26, 57, 58, 62].”
- WILDCAT: Weakly Supervised Learning of Deep ConvNets for Image Classification, Pointwise Localization and Segmentation (2017, unknown): “Recent alternatives include Global Average Pooling (GAP) [70], soft max in LSE pooling [58], Learning from Label Proportion (LLP) [65, 36], and top max scoring [39].”
- WELDON: Weakly Supervised Learning of Deep Convolutional Neural Networks (2016, methodology): “Recently, interesting MIL extensions have been introduced in [51, 24, 27, 38, 10].”
- Multiple instance learning: A survey of problem characteristics and applications (2016, methodology): “For example, in computer vision applications, it is difficult to guarantee that negative images contain no positive patches: An image showing a house may contain flowers, but is unlikely to be annotated as a flower image [92].”
- AdaScan: Adaptive Scan Pooling in Deep Convolutional Neural Networks for Human Action Recognition in Videos (2016, methodology): “Such solutions either use Latent Variable Models [22, 36, 9, 30, 19], which require an additional inference step during learning, or employ a variant of Recurrent Neural Networks (RNN) [29, 49] which have intermediate hidden states that are not immediately inter-pretable.”
- Weakly-supervised Video Anomaly Detection with Robust Temporal Feature Magnitude Learning (2021, methodology): “, we use `2 norm to compute the feature magnitude), where instead of assuming classification separability between normal and abnormal snippets (as assumed in [24]), we make a milder assumption that E[‖x‖2] ≥ E[‖x‖2].”