skip to main content

Title: Visually-Grounded Library of Behaviors for Manipulating Diverse Objects across Diverse Configurations and Views
We propose a visually-grounded library of behaviors approach for learning to manipulate diverse objects across varying initial and goal configurations and camera placements. Our key innovation is to disentangle the standard image-to-action mapping into two separate modules that use different types of perceptual input:(1) a behavior selector which conditions on intrinsic and semantically-rich object appearance features to select the behaviors that can successfully perform the desired tasks on the object in hand, and (2) a library of behaviors each of which conditions on extrinsic and abstract object properties, such as object location and pose, to predict actions to execute over time. The selector uses a semantically-rich 3D object feature representation extracted from images in a differential end-to-end manner. This representation is trained to be view-invariant and affordance-aware using self-supervision, by predicting varying views and successful object manipulations. We test our framework on pushing and grasping diverse objects in simulation as well as transporting rigid, granular, and liquid food ingredients in a real robot setup. Our model outperforms image-to-action mappings that do not factorize static and dynamic object properties. We further ablate the contribution of the selector's input and show the benefits of the proposed view-predictive, affordance-aware 3D visual object representations.
Authors:
; ; ; ; ; ;
Award ID(s):
1849287
Publication Date:
NSF-PAR ID:
10333940
Journal Name:
5th Annual Conference on Robot Learning
Sponsoring Org:
National Science Foundation
More Like this
  1. Vedaldi, Andrea ; Bischof, Horst ; Brox, Thomas ; Frahm, Jan-Michael (Ed.)
    Novel view video synthesis aims to synthesize novel viewpoints videos given input captures of a human performance taken from multiple reference viewpoints and over consecutive time steps. Despite great advances in model-free novel view synthesis, existing methods present three limitations when applied to complex and time-varying human performance. First, these methods (and related datasets) mainly consider simple and symmetric objects. Second, they do not enforce explicit consistency across generated views. Third, they focus on static and non-moving objects. The fine-grained details of a human subject can therefore suffer from inconsistencies when synthesized across different viewpoints or time steps. To tackle these challenges, we introduce a human-specific framework that employs a learned 3D-aware representation. Specifically, we first introduce a novel siamese network that employs a gating layer for better reconstruction of the latent volumetric representation and, consequently, final visual results. Moreover, features from consecutive time steps are shared inside the network to improve temporal consistency. Second, we introduce a novel loss to explicitly enforce consistency across generated views both in space and in time. Third, we present the Multi-View Human Action (MVHA) dataset, consisting of near 1200 synthetic human performance captured from 54 viewpoints. Experiments on the MVHA, Pose-Varying Human Modelmore »and ShapeNet datasets show that our method outperforms the state-of-the-art baselines both in view generation quality and spatio-temporal consistency.« less
  2. Human novel view synthesis aims to synthesize target views of a human subject given input images taken from one or more reference viewpoints. Despite significant advances in model-free novel view synthesis, existing methods present two major limitations when applied to complex shapes like humans. First, these methods mainly focus on simple and symmetric objects, e.g., cars and chairs, limiting their performances to fine-grained and asymmetric shapes. Second, existing methods cannot guarantee visual consistency across different adjacent views of the same object. To solve these problems, we present in this paper a learning framework for the novel view synthesis of human subjects, which explicitly enforces consistency across different generated views of the subject. Specifically, we introduce a novel multi-view supervision and an explicit rotational loss during the learning process, enabling the model to preserve detailed body parts and to achieve consistency between adjacent synthesized views. To show the superior performance of our approach, we present qualitative and quantitative results on the Multi-View Human Action (MVHA) dataset we collected (consisting of 3D human models animated with different Mocap sequences and captured from 54 different viewpoints), the Pose-Varying Human Model (PVHM) dataset, and ShapeNet. The qualitative and quantitative results demonstrate that our approachmore »outperforms the state-of-the-art baselines in both per-view synthesis quality, and in preserving rotational consistency and complex shapes (e.g. fine-grained details, challenging poses) across multiple adjacent views in a variety of scenarios, for both humans and rigid objects.« less
  3. We present a new weakly supervised learning-based method for generating novel category-specific 3D shapes from unoccluded image collections. Our method is weakly supervised and only requires silhouette annotations from unoccluded, category-specific objects. Our method does not require access to the object's 3D shape, multiple observations per object from different views, intra-image pixel correspondences, or any view annotations. Key to our method is a novel multi-projection generative adversarial network (MP-GAN) that trains a 3D shape generator to be consistent with multiple 2D projections of the 3D shapes, and without direct access to these 3D shapes. This is achieved through multiple discriminators that encode the distribution of 2D projections of the 3D shapes seen from a different views. Additionally, to determine the view information for each silhouette image, we also train a view prediction network on visualizations of 3D shapes synthesized by the generator. We iteratively alternate between training the generator and training the view prediction network. We validate our multi-projection GAN on both synthetic and real image datasets. Furthermore, we also show that multi-projection GANs can aid in learning other high-dimensional distributions from lower dimensional training datasets, such as material-class specific spatially varying reflectance properties from images.
  4. Drilling and milling operations are material removal processes involved in everyday conventional productions, especially in the high-speed metal cutting industry. The monitoring of tool information (wear, dynamic behavior, deformation, etc.) is essential to guarantee the success of product fabrication. Many methods have been applied to monitor the cutting tools from the information of cutting force, spindle motor current, vibration, as well as sound acoustic emission. However, those methods are indirect and sensitive to environmental noises. Here, the in-process imaging technique that can capture the cutting tool information while cutting the metal was studied. As machinists judge whether a tool is worn-out by the naked eye, utilizing the vision system can directly present the performance of the machine tools. We proposed a phase shifted strobo-stereoscopic method (Figure 1) for three-dimensional (3D) imaging. The stroboscopic instrument is usually applied for the measurement of fast-moving objects. The operation principle is as follows: when synchronizing the frequency of the light source illumination and the motion of object, the object appears to be stationary. The motion frequency of the target is transferring from the count information of the encoder signals from the working rotary spindle. If small differences are added to the frequency, the objectmore »appears to be slowly moving or rotating. This effect can be working as the source for the phase-shifting; with this phase information, the target can be whole-view 3D reconstructed by 360 degrees. The stereoscopic technique is embedded with two CCD cameras capturing images that are located bilateral symmetrically in regard to the target. The 3D scene is reconstructed by the location information of the same object points from both the left and right images. In the proposed system, an air spindle was used to secure the motion accuracy and drilling/milling speed. As shown in Figure 2, two CCDs with 10X objective lenses were installed on a linear rail with rotary stages to capture the machine tool bit raw picture for further 3D reconstruction. The overall measurement process was summarized in the flow chart (Figure 3). As the count number of encoder signals is related to the rotary speed, the input speed (unit of RPM) was set as the reference signal to control the frequency (f0) of the illumination of the LED. When the frequency was matched with the reference signal, both CCDs started to gather the pictures. With the mismatched frequency (Δf) information, a sequence of images was gathered under the phase-shifted process for a whole-view 3D reconstruction. The study in this paper was based on a 3/8’’ drilling tool performance monitoring. This paper presents the principle of the phase-shifted strobe-stereoscopic 3D imaging process. A hardware set-up is introduced, , as well as the 3D imaging algorithm. The reconstructed image analysis under different working speeds is discussed, the reconstruction resolution included. The uncertainty of the imaging process and the built-up system are also analyzed. As the input signal is the working speed, no other information from other sources is required. This proposed method can be applied as an on-machine or even in-process metrology. With the direct method of the 3D imaging machine vision system, it can directly offer the machine tool surface and fatigue information. This presented method can supplement the blank for determining the performance status of the machine tools, which further guarantees the fabrication process.« less
  5. We present MultiBodySync, a novel, end-to-end trainable multi-body motion segmentation and rigid registration framework for multiple input 3D point clouds. The two non-trivial challenges posed by this multi-scan multibody setting that we investigate are: (i) guaranteeing correspondence and segmentation consistency across multiple input point clouds capturing different spatial arrangements of bodies or body parts; and (ii) obtaining robust motion-based rigid body segmentation applicable to novel object categories. We propose an approach to address these issues that incorporates spectral synchronization into an iterative deep declarative network, so as to simultaneously recover consistent correspondences as well as motion segmentation. At the same time, by explicitly disentangling the correspondence and motion segmentation estimation modules, we achieve strong generalizability across different object categories. Our extensive evaluations demonstrate that our method is effective on various datasets ranging from rigid parts in articulated objects to individually moving objects in a 3D scene, be it single-view or full point clouds.