Objects occupy physical space and obey physical laws. To truly understand a scene, we must reason about the space that objects in it occupy, and how each objects is supported stably by each other. In other words, we seek to understand which objects would, if moved, cause other objects to fall. This 3D volumetric reasoning is important for many scene understanding tasks, ranging from segmentation of objects to perception of a rich 3D, physically well-founded, interpretations of the scene. In this paper, we propose a new algorithm to parse a single RGB-D image with 3D block units while jointly reasoning about the segments, volumes, supporting relationships, and object stability. Our algorithm is based on the intuition that a good 3D representation of the scene is one that fits the depth data well, and is a stable, self-supporting arrangement of objects (i.e., one that does not topple). We design an energy function for representing the quality of the block representation based on these properties. Our algorithm fits 3D blocks to the depth values corresponding to image segments, and iteratively optimizes the energy function. Our proposed algorithm is the first to consider stability of objects in complex arrangements for reasoning about the underlying structure of the scene. Experimental results show that our stability-reasoning framework improves RGB-D segmentation and scene volumetric representation.

3D volumetric reasoning is important for truly understanding a scene. Humans are able to both segment each object in an image, and perceive a rich 3D interpretation of the scene, e.g., the space an object occupies, which objects support other objects, and which objects would, if moved, cause other objects to fall. We propose a new approach for parsing RGB-D images using 3D block units for volumetric reasoning. The algorithm fits image segments with 3D blocks, and iteratively evaluates the scene based on block interaction properties. We produce a 3D representation of the scene based on jointly optimizing over segmentations, block fitting, supporting relations, and object stability. Our algorithm incorporates the intuition that a good 3D representation of the scene is the one that fits the data well, and is a stable, self-supporting (i.e., one that does not topple) arrangement of objects. We experiment on several datasets including controlled and real indoor scenarios. Results show that our stability-reasoning framework improves RGB-D segmentation and scene volumetric representation.

To enable high-level understanding of a scene, it is important to understand the occlusion and connected boundaries of objects in the image. In this paper, we propose a new framework for inferring boundaries from color and depth information. Even with depth information, it is not a trivial task to find and classify boundaries. Real-world depth images are noisy, especially at object boundaries, where our task is focused. Our approach uses features from both the color (which are sharp at object boundaries) and depth images (for providing geometric cues) to detect boundaries and classify them as occlusion or connected boundaries. We propose depth features based on surface fitting from sparse point clouds, and perform inference with a Conditional Random Field. One advantage of our approach is that occlusion and connected boundaries are identified with a single, common model. Experiments show that our mid-level color and depth features outperform using either depth or color alone, and our method surpasses the performance of baseline boundary detection methods.

Photography on a mobile camera provides access to additional sensors. In this paper, we estimate the absolute orientation of a planar object with respect to the ground, which can be a valuable prior for many vision tasks. To find the planar object orientation, our novel algorithm combines information from a gravity sensor with a planar homography that matches a region of an image to a training image (e.g., of a company logo). We demonstrate our approach with an iPhone application that records the gravity direction for each captured image. We find a homography that maps the training image to the test image, and propose a novel homography decomposition to extract the rotation matrix. We believe this is the first paper to estimate absolute planar object orientation by combining the inertial sensor information with vision algorithms. Experiments show that our proposed algorithm performs reliably.

Depth ordering is instrumental for understanding the 3D geometry of an image. Humans are surprisingly good at depth ordering even with abstract 2D line drawings. In this paper we propose a learning-based framework for depth ordering inference. Boundary and junction characteristics are important clues for this task, and we have developed new features based on these attributes. Although each feature individually can produce reasonable depth ordering results, each still has limitations, and we can achieve better performance by combining them. In practice, local depth ordering inferences can be contradictory. Therefore, we propose a Markov Random Field model with terms that are more global than previous work, and use graph optimization to encourage a globally consistent ordering. In addition, to produce better object segmentation for the task of depth ordering, we propose to explicitly enforce closed loops and long edges for the occlusion boundary detection. We collect a new depth-order dataset for this problem, including more than a thousand human-labeled images with various daily objects and configurations. The proposed algorithm shows promising performance over conventional methods on both synthetic and real scenes.

The most popular approach to large scale image retrieval is based on the bag-of-visual-word (BoV) representation of images. The spatial information is usually re-introduced as a post-processing step to re-rank the retrieved images, through a spatial verification like RANSAC. Since the spatial verification techniques are computationally expensive, they can be applied only to the top images in the initial ranking. In this paper, we propose an approach that can encode more spatial information into BoV representation and that is efficient enough to be applied to large-scale databases. Other works pursuing the same purpose have proposed exploring the word co-occurrences in the neighborhood areas. Our approach encodes more spatial information through the geometry-preserving visual phrases (GVP). In addition to co-occurrences, the GVP method also captures the local and long-range spatial layouts of the words. Our GVP based searching algorithm increases little memory usage or computational time compared to the BoV method. Moreover, we show that our approach can also be integrated to the min-hash method to improve its retrieval accuracy. The experiment results on Oxford 5K and Flicker 1M dataset show that our approach outperforms the BoV method even following a RANSAC verification.

A laser scan is useful in building the 3D model, and in one ran it can capture thousands of 3D points. However these 3D points are sparse compared to a normal image, which can easily have millions of pixels. To achieve a denser 3D map, 3D-interpolation is applied to each pixel in the image. In this work we propose an algorithm to combine the 3D geometry and the color for 3D-interpolation. We segment the 3D points based on their latent surfaces, and combine the surfaces with color through Markov Random Field. We find that the 3D geometry provides rich information for interpolation: 3D points with similar colors can be robustly clustered where not possible in the color space, and the interpolation can be performed on a better fitting surface rather than on the locally linear ones. Our experiments show that the proposed algorithm outperforms the baselines.

Object detection is a basic skill for a robot to perfor-m tasks in human environments. In order to build a good object classifier, a large training set of la-beled images is required; this is typically collect-ed and labeled (often painstakingly) by a human. This method is not scalable and therefore limits the robot’s detection performance. We propose an algorithm for a robot to collect more data in the environment during its training phase so that in the future it could detect objects more reli-ably. The first step is to plan a path for collecting additional training images, which is hard because a previously visited location affects the decision for the future locations. One key component of our work is path planning by building a sparse graph that captures these dependencies. The other key component is our learning algorithm that weighs the errors made in robot’s data collection process while updating the classifier. In our experiments, we show that our algorithms enable the robot to im-prove its object classifiers significantly.

10.5244/C.24.46