Strong supervision from weak annotation: Interactive training of deformable part models

By Steve Branson, Pietro Perona, Serge J. Belongie. 2011

115 citations recorded by the source corpus.

Subfields: Aerospace Engineering, 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 Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Interpreting CNNs via Decision Trees (2018, background): “Like in [3, 28], we ignored species labels and regarded all these images as a single bird category.”
Augmenting Strong Supervision Using Web Data for Fine-Grained Categorization (2015, methodology): “Our method is also related to the strategy of training with auxiliary data sources, such as domain adaptation [16] for heterogeneous data sources and incremental learning [6, 30] for homogeneous ones.”
Leveraging the Wisdom of the Crowd for Fine-Grained Recognition (2016, unknown): “Fine-grained recognition is challenging.”
Detect What You Can: Detecting and Representing Objects Using Holistic Models and Body Parts (2014, methodology): “Similar to our method, [1, 23, 3] use additional annotations to provide supervision for a part-based model, while Poselets [2] and [14] use keypoints and masks annotations to cluster objects or their parts.”
Object Detection Using Strongly-Supervised Deformable Part Models (2012, unknown): “The task of part localization addressed in [10, 16, 24], however, is different from the task of object detection in this paper.”
Local Alignments for Fine-Grained Categorization (2014, methodology): “semantic parts that are shared by all (or at least most of) the sub-classes, as in (Branson et al., 2011; Liu et al., 2012), such as the head of a dog or the body of a bird, as opposed to distinctive parts, as in (Yao et al.”
Y Qualification Main task Y Return N N Y User qualified ? N Feedback Pass ? Pass ? Annotator training Annotating images (2017, methodology): “Click annotation schemes have been used in part-based detection to annotate part locations of an object [9, 74], and in human pose estimation to annotate key-points of human body parts [26, 49, 56].”
A scalable active framework for region annotation in 3D shape collections (2016, background): “Their main advantage is that they acquire training data incrementally, enabling a classifier to predict which samples need to be labeled next to improve their performance [Vijayanarasimhan and Grauman 2008; Branson et al. 2011; Vezhnevets et al. 2012; Branson et al. 2014].”
Low-Rank Bilinear Pooling for Fine-Grained Classification (2016, background): “However, such supervised annotations are costly to obtain, leading to efforts to utilize interactive learning [2] or partially supervised discovery of discriminative patches representing semantic parts [3].”
Embedding Human Knowledge into Deep Neural Network via Attention Map (2019, background): “One of the major approaches to embedding human knowledge into machine learning models is human-in-the-loop (HITL) (Branson et al., 2010; Deng et al., 2013; Branson et al., 2011; Parkash and Parikh, 2012; Parikh and Grauman, 2011; Duan et al., 2012).”
Nonparametric Part Transfer for Fine-Grained Recognition (2014, background): “Another line of work uses active classification techniques, which require human interaction during testing to refine classification results [35, 5].”
The Ignorant Led by the Blind: A Hybrid Human–Machine Vision System for Fine-Grained Categorization (2014, methodology): “A number of exciting active learning algorithms that incorporate new types of human interactivity have come about in recent years (Vijayanarasimhan and Grauman 2009, 2011; Donahue and Grauman 2011; Vondrick and Ramanan 2011; Settles 2008; Parkash and Parikh 2012; Branson et al. 2011).”
Inferring Generative Model Structure with Static Analysis (2017, background): “attributes [18, 19, 29], like location and size, to weakly supervise more complex image-based learning tasks [5, 7, 11, 26, 38].”
Deformable Part Descriptors for Fine-Grained Recognition and Attribute Prediction (2013, methodology): “Some techniques use humans-in-the-loop, asking human annotators to click on object parts, answer questions regarding object attributes [13], or mark the region that best differentiates two confusing categories [17].”
Mining Object Parts from CNNs via Active Question-Answering (2017, methodology): “[4] used human-computer interactions to label object parts to learn part models.”