|Abstract: ||Object recognition and pose estimation is a fundamental problem in computer vision and of utmost importance in robotic applications. Object recognition refers to the problem of recognizing certain object instances, or categorizing objects into specific classes. Pose estimation deals with estimating the exact position of the object in 3D space, usually expressed in Euler angles. There are generally two types of objects that require special care when designing solutions to the aforementioned problems: rigid and deformable. Dealing with deformable objects has been a much harder problem, and usually solutions that apply to rigid objects, fail when used for deformable objects due to the inherent assumptions made during the design.
In this thesis we deal with object categorization, instance recognition and pose estimation of both rigid and deformable objects. In particular, we are interested in a special type of deformable objects, clothes. We tackle the problem of autonomously recognizing and unfolding articles of clothing using a dual manipulator. This problem consists of grasping an article from a random point, recognizing it and then bringing it into an unfolded state by a dual arm robot. We propose a data-driven method for clothes recognition from depth images using Random Decision Forests. We also propose a method for unfolding an article of clothing after estimating and grasping two key-points, using Hough Forests. Both methods are implemented into a POMDP framework allowing the robot to interact optimally with the garments, taking into account uncertainty in the recognition and point estimation process. This active recognition and unfolding makes our system very robust to noisy observations. Our methods were tested on regular-sized clothes using a dual-arm manipulator. Our systems perform better in both accuracy and speed compared to state-of-the-art approaches.
In order to take advantage of the robotic manipulator and increase the accuracy of our system, we developed a novel approach to address generic active vision problems, called Active Random Forests. While state of the art focuses on best viewing parameters selection based on single view classifiers, we propose a multi-view classifier where the decision mechanism of optimally changing viewing parameters is inherent to the classification process. This has many advantages: a) the classifier exploits the entire set of captured images and does not simply aggregate probabilistically per view hypotheses; b) actions are based on learnt disambiguating features from all views and are optimally selected using the powerful voting scheme of Random Forests and c) the classifier can take into account the costs of actions. The proposed framework was applied to the same task of autonomously unfolding clothes by a robot, addressing the problem of best viewpoint selection in classification, grasp point and pose estimation of garments. We show great performance improvement compared to state of the art methods and our previous POMDP formulation.
Moving from deformable to rigid objects while keeping our interest to domestic robotic applications, we focus on object instance recognition and 3D pose estimation of household objects. We are particularly interested in realistic scenes that are very crowded and objects can be perceived under severe occlusions. Single shot-based 6D pose estimators with manually designed features are still unable to tackle such difficult scenarios for a variety of objects, motivating the research towards unsupervised feature learning and next-best-view estimation. We present a complete framework for both single shot-based 6D object pose estimation and next-best-view prediction based on Hough Forests, the state of the art object pose estimator that performs classification and regression jointly. Rather than using manually designed features we propose an unsupervised feature learnt from depth-invariant patches using a Sparse Autoencoder. Furthermore, taking advantage of the clustering performed in the leaf nodes of Hough Forests, we learn to estimate the reduction of uncertainty in other views, formulating the problem of selecting the next-best-view. To further improve 6D object pose estimation, we propose an improved joint registration and hypotheses verification module as a final refinement step to reject false detections. We provide two additional challenging datasets inspired from realistic scenarios to extensively evaluate the state of the art and our framework. One is related to domestic environments and the other depicts a bin-picking scenario mostly found in industrial settings. We show that our framework significantly outperforms state of the art both on public and on our datasets.
Unsupervised feature learning, although efficient, might produce sub-optimal features for our particular tast. Therefore in our last work, we leverage the power of Convolutional Neural Networks to tackled the problem of estimating the pose of rigid objects by an end-to-end deep regression network. To improve the moderate performance of the standard regression objective function, we introduce the Siamese Regression Network. For a given image pair, we enforce a similarity measure between the representation of the sample images in the feature and pose space respectively, that is shown to boost regression performance. Furthermore, we argue that our pose-guided feature learning using our Siamese Regression Network generates more discriminative features that outperform the state of the art. Last, our feature learning formulation provides the ability of learning features that can perform under severe occlusions, demonstrating high performance on our novel hand-object dataset.
Concluding, this work is a research on the area of object detection and pose estimation in 3D space, on a variety of object types. Furthermore we investigate how accuracy can be further improved by applying active vision techniques to optimally move the camera view to minimize the detection error.|