Laura Antanas

CHARISMA: An integrated approach to automatic H&E-stained skeletal muscle cell segmentation using supervised learning and novel robust clump splitting

Histological image analysis plays a key role in understanding the effects of disease and treatment responses at the cellular level. However, evaluating histology images by hand is time-consuming and subjective. While semi-automatic and automatic approaches for image segmentation give acceptable results in some branches of histological image analysis, until now this has not been the case when applied to skeletal muscle histology images. We introduce Charisma, a new top-down cell segmentation framework for histology images which combines image processing techniques, a supervised trained classifier and a novel robust clump splitting algorithm. We evaluate our framework on real-world data from intensive care unit patients. Considering both segmentation and cell property distributions, the results obtained by our method correspond well to the ground truth, outperforming other examined methods.

A relational kernel-based framework for hierarchical image understanding

While relational representations have been popular in early work on syntactic and structural pattern recognition, they are rarely used in contemporary approaches to computer vision due to their pure symbolic nature. The recent progress and successes in combining statistical learning principles with relational representations motivates us to reinvestigate the use of such representations. More specifically, we show that statistical relational learning can be successfully used for hierarchical image understanding. We employ kLog, a new logical and relational language for learning with kernels to detect objects at different levels in the hierarchy. The key advantage of kLog is that both appearance features and rich, contextual dependencies between parts in a scene can be integrated in a principled and interpretable way to obtain a qualitative representation of the problem. At each layer, qualitative spatial structures of parts in images are detected, classified and then employed one layer up the hierarchy to obtain higher-level semantic structures. We apply a four-layer hierarchy to street view images and successfully detect corners, windows, doors, and individual houses.

There are plenty of places like home: Using relational representations in hierarchies for distance-based image understanding

Understanding images in terms of logical and hierarchical structures is crucial for many semantic tasks, including image retrieval, scene understanding and robotic vision. This paper combines robust feature extraction, qualitative spatial relations, relational instance-based learning and compositional hierarchies in one framework. For each layer in the hierarchy, qualitative spatial structures in images are detected, classified and then employed one layer up the hierarchy to obtain higher-level semantic structures. We apply a four-layer hierarchy to street view images and subsequently detect corners, windows, doors, and individual houses.

A relational kernel-based approach to scene classification

Real-world scenes involve many objects that interact with each other in complex semantic patterns. For example, a bar scene can be naturally described as having a variable number of chairs of similar size, close to each other and aligned horizontally. This high-level interpretation of a scene relies on semantically meaningful entities and is most generally described using relational representations or (hyper-) graphs. Popular in early work on syntactic and structural pattern recognition, relational representations are rarely used in computer vision due to their pure symbolic nature. Yet, today recent successes in combining them with statistical learning principles motivates us to reinvestigate their use. In this paper we show that relational techniques can also improve scene classification. More specifically, we employ a new relational language for learning with kernels, called kLog. With this language we define higher-order spatial relations among semantic objects. When applied to a particular image, they characterize a particular object arrangement and provide discriminative cues for the scene category. The kernel allows us to tractably learn from such complex features. Thus, our contribution is a principled and interpretable approach to learn from symbolic relations how to classify scenes in a statistical framework. We obtain results comparable to state-of-the-art methods on 15 Scenes and a subset of the MIT indoor dataset.

Opening doors: An initial SRL approach

Opening doors is an essential task that a robot should perform. In this paper, we propose a logical approach to predict the action of opening doors, together with the action point where the action should be performed. The input of our system is a pair of bounding boxes of the door and door handle, together with background knowledge in the form of logical rules. Learning and inference are performed with the probabilistic programming language ProbLog. We evaluate our approach on a doors dataset and we obtain encouraging results. Additionally, a comparison to a propositional decision tree shows the benefits of using a probabilistic programming language such as ProbLog.

Relational affordance learning for task-dependent robot grasping

Robot grasping depends on the specific manipulation scenario: the object, its properties, task and grasp constraints. Object-task affordances facilitate semantic reasoning about pre-grasp configurations with respect to the intended tasks, favoring good grasps. We employ probabilistic rule learning to recover such object-task affordances for task-dependent grasping from realistic video data.

Relational Symbol Grounding through Affordance Learning: An Overview of the ReGround Project

Symbol grounding is the problem of associating symbols from language with a corresponding referent in the environment. Traditionally, research has focused on identifying single objects and their properties. The ReGround project hypothesizes that the grounding process must consider the full context of the environment, including multiple objects, their properties, and relationships among these objects. ReGround targets the development of a novel framework for “affordance grounding”, by which an agent placed in a new environment can adapt to its new setting and interpret possibly multi-modal input in order to correctly carry out the requested tasks.

Relational Kernel-Based Grasping with Numerical Features

Object grasping is a key task in robot manipulation. Performing a grasp largely depends on the object properties and grasp constraints. This paper proposes a new statistical relational learning approach to recognize graspable points in object point clouds. We characterize each point with numerical shape features and represent each cloud as a (hyper-) graph by considering qualitative spatial relations between neighboring points. Further, we use kernels on graphs to exploit extended contextual shape information and compute discriminative features which show improvement upon local shape features. Our work for robot grasping highlights the importance of moving towards integrating relational representations with low-level descriptors for robot vision. We evaluate our relational kernel-based approach on a realistic dataset with 8 objects.

Intra-patient Non-rigid Registration of 3D Vascular Cerebral Images

Accuracy of intensity-based non-rigid registration of vascular images is deteriorated in the presence of excessive noise as occurring in CBCT angiography. A new approach to non-rigid registration of vascular images is presented based on the assumption of isometric deformation of vessel structures. For every voxel in the vascular image, the distances along the vessels to some reference voxels are computed as features that are invariant under isometric deformation. Due to the global information on vessel connectivity that is encoded in these features, voxel-based registration of these feature images can compensate for local optima in direct intensity-based registration of the vascular images. The method is validated in the context of brain-shift mitigation, but the technique can be used more broadly. Tests on artificially deformed vascular images show that our algorithm reaches higher accuracy than traditional intensity-based registration.

Not far away from home: A relational distance-based approach to understand images of houses

This book constitutes the thoroughly refereed post-proceedings of the 20th International Conference on Inductive Logic Programming, ILP 2010, held in Florence, Italy in June 2010. The 11 revised full papers and 15 revised short papers presented together with abstracts of three invited talks were carefully reviewed and selected during two rounds of refereeing and revision. All current issues in inductive logic programming, i.e. in logic programming for machine learning are addressed, in particular statistical learning and other probabilistic approaches to machine learning are reflected.