179,176 research outputs found
Automatic annotation for weakly supervised learning of detectors
No full textPhDObject detection in images and action detection in videos are among the most widely studied
computer vision problems, with applications in consumer photography, surveillance, and automatic
media tagging. Typically, these standard detectors are fully supervised, that is they require
a large body of training data where the locations of the objects/actions in images/videos have
been manually annotated. With the emergence of digital media, and the rise of high-speed internet,
raw images and video are available for little to no cost. However, the manual annotation
of object and action locations remains tedious, slow, and expensive. As a result there has been
a great interest in training detectors with weak supervision where only the presence or absence
of object/action in image/video is needed, not the location. This thesis presents approaches for
weakly supervised learning of object/action detectors with a focus on automatically annotating
object and action locations in images/videos using only binary weak labels indicating the presence
or absence of object/action in images/videos.
First, a framework for weakly supervised learning of object detectors in images is presented.
In the proposed approach, a variation of multiple instance learning (MIL) technique for automatically
annotating object locations in weakly labelled data is presented which, unlike existing
approaches, uses inter-class and intra-class cue fusion to obtain the initial annotation. The initial
annotation is then used to start an iterative process in which standard object detectors are used to
refine the location annotation. Finally, to ensure that the iterative training of detectors do not drift
from the object of interest, a scheme for detecting model drift is also presented. Furthermore,
unlike most other methods, our weakly supervised approach is evaluated on data without manual
pose (object orientation) annotation.
Second, an analysis of the initial annotation of objects, using inter-class and intra-class cues,
is carried out. From the analysis, a new method based on negative mining (NegMine) is presented
for the initial annotation of both object and action data. The NegMine based approach is a
much simpler formulation using only inter-class measure and requires no complex combinatorial
optimisation but can still meet or outperform existing approaches including the previously pre3
sented inter-intra class cue fusion approach. Furthermore, NegMine can be fused with existing
approaches to boost their performance.
Finally, the thesis will take a step back and look at the use of generic object detectors as prior
knowledge in weakly supervised learning of object detectors. These generic object detectors are
typically based on sampling saliency maps that indicate if a pixel belongs to the background
or foreground. A new approach to generating saliency maps is presented that, unlike existing
approaches, looks beyond the current image of interest and into images similar to the current
image. We show that our generic object proposal method can be used by itself to annotate the
weakly labelled object data with surprisingly high accuracy
Supervised learning of arithmetic invariants
No full textWe survey some recent implementations of supervised learning techniques on large sets of arithmetic data. As part of our methodological review, we perform some rudimentary statistical learning algorithms by hand on simplified problems. We incorporate a self-contained number-theoretic background which places a significant emphasis on conjectures and examples relevant to the machine learning context
Weakly Supervised Learning of Objects, Attributes and their Associations
No full textAbstract. When humans describe images they tend to use combinations of nouns and adjectives, corresponding to objects and their associated attributes respectively. To generate such a description automatically, one needs to model objects, attributes and their associations. Conventional methods require strong annotation of object and attribute locations, making them less scalable. In this paper, we model objectattribute associations from weakly labelled images, such as those widely available on media sharing sites (e.g. Flickr), where only image-level labels (either object or attributes) are given, without their locations and associations. This is achieved by introducing a novel weakly supervised non-parametric Bayesian model. Once learned, given a new image, our model can describe the image, including objects, attributes and their associations, as well as their locations and segmentation. Extensive experiments on benchmark datasets demonstrate that our weakly supervised model performs at par with strongly supervised models on tasks such as image description and retrieval based on object-attribute associations
A semi-supervised spam mail detector
No full textThis document describes a novel semi-supervised approach to spam classification, which was successful at the ECML/PKDD 2006 spam classification challenge. A local learning method based on lazy projections was successfully combined with a variant of a standard semi-supervised learning algorithm
Efficient Non-Parametric Function Induction in Semi-Supervised Learning
No full textThere has been an increase of interest for semi-supervised learning recently, because of the many datasets with large amounts of unlabeled examples and only a few labeled ones. This paper follows up on proposed non-parametric algorithms which provide an estimated continuous label for the given unlabeled examples. It extends them to function induction algorithms that correspond to the minimization of a regularization criterion applied to an out-of-sample example, and happens to have the form of a Parzen windows regressor. The advantage of the extension is that it allows predicting the label for a new example without having to solve again a linear system of dimension 'n' (the number of unlabeled and labeled training examples), which can cost O(n^3). Experiments show that the extension works well, in the sense of predicting a label close to the one that would have been obtained if the test example had been included in the unlabeled set. This relatively efficient function induction procedure can also be used when 'n' is large to approximate the solution by writing it only in terms of a kernel expansion with 'm' Il y a eu un regain d'intérêt récemment pour l'apprentissage semi-supervisé, à cause du grand nombre de bases de données comportant de très nombreux exemples non étiquetés et seulement quelques exemples étiquetés. Cet article poursuit le travail fait sur les algorithmes non paramétriques qui fournissent une étiquette continue estimée pour les exemples non-étiquetés. Il les étend à des algorithmes d'induction fonctionnelle qui correspondent à la minimisation d'un critère de régularisation appliqué à un exemple hors-échantillon, et qui ont la forme d'un régresseur à fenêtre de Parzen. L'avantage de cette extension est qu'elle permet de prédire l'étiquette d'un nouvel exemple sans avoir à résoudre de nouveau un système de dimension 'n' (le nombre d'exemples d'entraînement total), qui peut être de l'ordre de O(n^3). Les expériences montrent que l'extension fonctionne bien, en ce sens que l'étiquette prédite est proche de celle qui aurait été obtenue si l'exemple de test avait fait partie de l'ensemble non étiqueté. Cette procédure d'induction fonctionnelle relativement efficace peut également être utilisée, lorsque 'n' est grand, pour estimer la solution en l'écrivant seulement en fonction d'une expansion à noyau avec 'm'non-parametric models, classification, regression, semi-supervised learning, modèles non paramétriques, classification, régression, apprentissage semi-supervisé
Self-supervised learning for transferable representations
No full textMachine learning has undeniably achieved remarkable advances thanks to large labelled datasets and supervised learning. However, this progress is constrained by the labour-intensive annotation process. It is not feasible to generate extensive labelled datasets for every problem we aim to address. Consequently, there has been a notable shift in recent times toward approaches that solely leverage raw data. Among these, self-supervised learning has emerged as a particularly powerful approach, offering scalability to massive datasets and showcasing considerable potential for effective knowledge transfer. This thesis investigates self-supervised representation learning with a strong focus on computer vision applications. We provide a comprehensive survey of self-supervised methods across various modalities, introducing a taxonomy that categorises them into four distinct families while also highlighting practical considerations for real-world implementation. Our focus thenceforth is on the computer vision modality, where we perform a comprehensive benchmark evaluation of state-of-the-art self supervised models against many diverse downstream transfer tasks. Our findings reveal that self-supervised models often outperform supervised learning across a spectrum of tasks, albeit with correlations weakening as tasks transition beyond classification, particularly for datasets with distribution shifts. Digging deeper, we investigate the influence of data augmentation on the transferability of contrastive learners, uncovering a trade-off between spatial and appearance-based invariances that generalise to real-world transformations. This begins to explain the differing empirical performances achieved by self-supervised learners on different downstream tasks, and it showcases the advantages of specialised representations produced with tailored augmentation. Finally, we introduce a novel self-supervised pre-training algorithm for object detection, aligning pre-training with downstream architecture and objectives, leading to reduced localisation errors and improved label efficiency. In conclusion, this thesis contributes a comprehensive understanding of self-supervised representation learning and its role in enabling effective transfer across computer vision tasks
Semi-supervised Learning for Anomalous Trajectory Detection
No full textA novel learning framework is proposed for anomalous behaviour detection in a video surveillance scenario, so that a classifier which distinguishes between normal and anomalous behaviour patterns can be incrementally trained with the assistance of a human operator. We consider the behaviour of pedestrians in terms of motion trajectories, and parametrise these trajectories using the control points of approximating cubic spline curves. This paper demonstrates an incremental semi-supervised one-class learning procedure in which unlabelled trajectories are combined with occasional examples of normal behaviour labelled by a human operator. This procedure is found to be effective on two different datasets, indicating that a human operator could potentially train the system to detect anomalous behaviour by providing only occasional interventions (a small percentage of the total number of observations)
A Taxonomy for Semi-Supervised Learning Methods
No full textWe propose a simple taxonomy of probabilistic graphical models for the semi-supervised learning problem. We give some broad classes of algorithms for each of the families and point to specific realizations in the literature. Finally, we shed more detailed light on the family of methods using input-dependent regularization (or conditional prior distributions) and show parallels to the Co-training paradigm.LAPMA
Neural Flocking: MPC-based Supervised Learning of Flocking Controllers
No full textWe show how a symmetric and fully distributed flocking controller
can be synthesized using Deep Learning from a centralized flocking
controller. Our approach is based on Supervised Learning, with the centralized
controller providing the training data, in the form of trajectories
of state-action pairs. We use Model Predictive Control (MPC) for the centralized
controller, an approach that we have successfully demonstrated
on flocking problems. MPC-based flocking controllers are high-performing
but also computationally expensive. By learning a symmetric and distributed
neural flocking controller from a centralized MPC-based one,
we achieve the best of both worlds: the neural controllers have high
performance (on par with the MPC controllers) and high efficiency. Our
experimental results demonstrate the sophisticated nature of the distributed
controllers we learn. In particular, the neural controllers are
capable of achieving myriad flocking-oriented control objectives, including
flocking formation, collision avoidance, obstacle avoidance, predator
avoidance, and target seeking. Moreover, they generalize the behavior
seen in the training data to achieve these objectives in a significantly
broader range of scenarios. In terms of verification of our neural flocking
controller, we use a form of statistical model checking to compute
confidence intervals for its convergence rate and time to convergence
Also By The Same Author: AKTiveAuthor, a Citation Graph Approach to Name Disambiguation
No full textThe desire for definitive data and the semantic web drive for inference over heterogeneous data sources requires co-reference resolution to be performed on those data. In particular, name disambiguation is required to allow accurate publication lists, citation counts and impact measures to be determined. This paper describes a graph-based approach to author disambiguation on large-scale citation networks. Using self-citation, co-authorship and document source analyses, AKTiveAuthor clusters papers, achieving precision of 0.997 and recall of 0.818 over a test group of eight surname clusters
- …
