Deep multiple instance learning for image classification and auto-annotation

By Jiajun Wu, Yinan Yu, Chang Huang, Kai Yu. 2015

429 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

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].”
TI-POOLING: Transformation-Invariant Pooling for Feature Learning in Convolutional Neural Networks (2016, background): “Convolutional neural networks are usually represented as a sequence of convolutional and subsampling layers with one or more fully-connected layers before the outputs.”
D3TW: Discriminative Differentiable Dynamic Time Warping for Weakly Supervised Action Alignment and Segmentation (2019, methodology): “For images, weakly supervised learning has been studied in classification [37, 24], semantic segmentation [40], object detection [22], and visual grounding [17, 38].”
Audio-Visual Event Localization in Unconstrained Videos (2018, positive): “only video-level labels are available, we infer label of each audio-visual segment pair in the training phase, and aggregate these individual predictions into a video-level prediction by MIL pooling as in [55]:”
Global Weighted Average Pooling Bridges Pixel-level Localization and Image-level Classification (2018, methodology): “Various WSL methods [4], [5], [6] demonstrate their potential of auto-localization to reduce the annotation effort.”
Unsupervised Joint Mining of Deep Features and Image Labels for Large-Scale Radiology Image Categorization and Scene Recognition (2017, methodology): “Image Categorization or Auto-annotation: Image auto-annotation task is addressed via multiple instance learning [59] but the target domain is restricted to a small subset (only 25 out of 1000 classes) of ImageNet [12] and SUN [61].”
Action Sets: Weakly Supervised Action Segmentation Without Ordering Constraints (2017, methodology): “We also evaluated the neural network based multiple instance learning approach of [37], which also was not able to make reliable predictions (accuracy 0.”
An In-field Automatic Wheat Disease Diagnosis System (2017, methodology): “To reduce the laborious manual annotation as much as possible and achieve object localization simultaneously, some works are focused on the combination of deep learning and MIL framework (Pinheiro and Collobert, 2015; Wu et al., 2015).”
Multiple instance learning: A survey of problem characteristics and applications (2016, background): “The weak supervision can come from description sentences [103–105], web search engine results [106], tags associated with similar images and words found on web pages associated with the images [2].”
Min-Entropy Latent Model for Weakly Supervised Object Detection (2018, background): “With the rise of deep learning, pipelined approaches have been evolving into multiple instance learning (MIL) networks [22]–[25], [27]–[29], [33]–[38].”
LEAP: Learning to Prescribe Effective and Safe Treatment Combinations for Multimorbidity (2017, methodology): “More recently, deep learning methods have been developed for multi-instance learning [37, 38].”
Multi-view Adversarially Learned Inference for Cross-domain Joint Distribution Matching (2018, background): “, (1) in image translation/editing [8, 16, 21, 30, 32], the goal is to transform an input image into an output one with desired properties; (2) in image annotating/attribute predicting [4, 5, 28, 29], one aims to extract meaningful semantic labels as many as possible from a given image; (3) in attri”
Learning to Separate Object Sounds by Watching Unlabeled Video (2018, neutral): “Max pooling in deep MIL [24,80,81] is typically used to identify the positive instances within an aggregated bag.”
Weakly Supervised Object Localization and Detection: A Survey (2021, neutral): “regularization [6] on the classic MIL models, such as LSVM [190], while the current research obtains the breakthrough by building deep MIL models [172], [179], [221].”
Unified Visual-Semantic Embeddings: Bridging Vision and Language With Structured Meaning Representations (2019, background): “The losses are weighted by the matching score, thus reinforce the correspondence between u + o and the matched region. softminpooling s local = 0.26 This technique is related to multi-instance learning [45].”
Unsupervised Visual Representation Learning by Graph-Based Consistent Constraints (2016, background): “Recent research on visual representation learning has been explored in a weakly supervised [15,16,17,18], semi-supervised [19,20] and unsupervised [11,12] fashion.”
Finding "It": Weakly-Supervised Reference-Aware Visual Grounding in Instructional Videos (2018, methodology): “MIL has been a effective framework for weakly-supervised learning in several applications, including image classification [50], object localization [13], tracking [5], and instance segmentation [37].”