Neural Assembler: Learning to Generate Fine-Grained Robotic Assembly Instructions from Multi-View Images

Image-guided object assembly represents a burgeoning research topic in computer vision. This paper introduces a novel task: translating multi-view images of a structural 3D model (for example, one constructed with building blocks drawn from a 3D-object library) into a detailed sequence of assembly instructions executable by a robotic arm. Fed with multi-view images of the target 3D model for replication, the model designed for this task must address several sub-tasks, including recognizing individual components used in constructing the 3D model, estimating the geometric pose of each component, and deducing a feasible assembly order adhering to physical rules. Establishing accurate 2D-3D correspondence between multi-view images and 3D objects is technically challenging. To tackle this, we propose an end-to-end model known as the Neural Assembler. This model learns an object graph where each vertex represents recognized components from the images, and the edges specify the topology of the 3D model, enabling the derivation of an assembly plan. We establish benchmarks for this task and conduct comprehensive empirical evaluations of Neural Assembler and alternative solutions. Our experiments clearly demonstrate the superiority of Neural Assembler.

Iterative Reconstruction Based on Latent Diffusion Model for Sparse Data Reconstruction

Reconstructing Computed tomography (CT) images from sparse measurement is a well-known ill-posed inverse problem. The Iterative Reconstruction (IR) algorithm is a solution to inverse problems. However, recent IR methods require paired data and the approximation of the inverse projection matrix. To address those problems, we present Latent Diffusion Iterative Reconstruction (LDIR), a pioneering zero-shot method that extends IR with a pre-trained Latent Diffusion Model (LDM) as a accurate and efficient data prior. By approximating the prior distribution with an unconditional latent diffusion model, LDIR is the first method to successfully integrate iterative reconstruction and LDM in an unsupervised manner. LDIR makes the reconstruction of high-resolution images more efficient. Moreover, LDIR utilizes the gradient from the data-fidelity term to guide the sampling process of the LDM, therefore, LDIR does not need the approximation of the inverse projection matrix and can solve various CT reconstruction tasks with a single model. Additionally, for enhancing the sample consistency of the reconstruction, we introduce a novel approach that uses historical gradient information to guide the gradient. Our experiments on extremely sparse CT data reconstruction tasks show that LDIR outperforms other state-of-the-art unsupervised and even exceeds supervised methods, establishing it as a leading technique in terms of both quantity and quality. Furthermore, LDIR also achieves competitive performance on nature image tasks. It is worth noting that LDIR also exhibits significantly faster execution times and lower memory consumption compared to methods with similar network settings. Our code will be publicly available.

FBNet: Feedback Network for Point Cloud Completion

The rapid development of point cloud learning has driven point cloud completion into a new era. However, the information flows of most existing completion methods are solely feedforward, and high-level information is rarely reused to improve low-level feature learning. To this end, we propose a novel Feedback Network (FBNet) for point cloud completion, in which present features are efficiently refined by rerouting subsequent fine-grained ones. Firstly, partial inputs are fed to a Hierarchical Graph-based Network (HGNet) to generate coarse shapes. Then, we cascade several Feedback-Aware Completion (FBAC) Blocks and unfold them across time recurrently. Feedback connections between two adjacent time steps exploit fine-grained features to improve present shape generations. The main challenge of building feedback connections is the dimension mismatching between present and subsequent features. To address this, the elaborately designed point Cross Transformer exploits efficient information from feedback features via cross attention strategy and then refines present features with the enhanced feedback features. Quantitative and qualitative experiments on several datasets demonstrate the superiority of proposed FBNet compared to state-of-the-art methods on point completion task.