The progression in large-scale 3D generative models has been impeded by significant resource requirements for training and challenges like inefficient representations. This paper introduces Make-A-Shape, a novel 3D generative model trained on a vast scale, using 10 million publicly-available shapes. We first innovate the wavelet-tree representation to encode high-resolution SDF shapes with minimal loss, leveraging our newly-proposed subband coefficient filtering scheme. We then design a subband coefficient packing scheme to facilitate diffusion-based generation and a subband adaptive training strategy for effective training on the large-scale dataset. Our generative framework is versatile, capable of conditioning on various input modalities such as images, point clouds, and voxels, enabling a variety of downstream applications, e.g., unconditional generation, completion, and conditional generation. Our approach clearly surpasses the existing baselines in delivering high-quality results and can efficiently generate shapes within two seconds for most conditions.

6-DoF pose estimation is an essential component of robotic manipulation pipelines. However, it usually suffers from a lack of generalization to new instances and object types. Most widely used methods learn to infer the object pose in a discriminative setup where the model filters useful information to infer the exact pose of the object. While such methods offer accurate poses, the model does not store enough information to generalize to new objects. In this work, we address the generalization capability of pose estimation using models that contain enough information about the object to render it in different poses. We follow the line of work that inverts neural renderers to infer the pose. We propose i-σSRN to maximize the information flowing from the input pose to the rendered scene and invert them to infer the pose given an input image. Specifically, we extend Scene Representation Networks (SRNs) by incorporating a separate network for density estimation and introduce a new way of obtaining a weighted scene representation. We investigate several ways of initial pose estimates and losses for the neural renderer. Our final evaluation shows a significant improvement in inference performance and speed compared to existing approaches.

Generating shapes using natural language can enable new ways of imagining and creating the things around us. While significant recent progress has been made in text-to-image generation, text-to-shape generation remains a challenging problem due to the unavailability of paired text and shape data at a large scale. We present a simple yet effective method for zero-shot text-to-shape gener-ation that circumvents such data scarcity. Our proposed method, named CLIP-Forge, is based on a two-stage training process, which only depends on an unlabelled shape dataset and a pre-trained image-text network such as CLIP. Our method has the benefits of avoiding expensive inference time optimization, as well as the ability to generate multiple shapes for a given text. We not only demonstrate promising zero-shot generalization of the CLIP-Forge model qualitatively and quantitatively, but also provide extensive compar-ative evaluations to better understand its behavior.

Reverse Engineering a CAD shape from other representations is an important geometric processing step for many downstream applications. In this work, we introduce a novel neural network architecture to solve this challenging task and approximate a smoothed signed distance function with an editable, constrained, prismatic CAD model. During training, our method reconstructs the input geometry in the voxel space by decomposing the shape into a series of 2D profile images and 1D envelope functions. These can then be recombined in a differentiable way allowing a geometric loss function to be defined. During inference, we obtain the CAD data by first searching a database of 2D constrained sketches to find curves which approximate the profile images, then extrude them and use Boolean operations to build the final CAD model. Our method approximates the target shape more closely than other methods and outputs highly editable constrained parametric sketches which are compatible with existing CAD software.

In this paper, we study how to determine concurrent transmissions and the transmission power level of each link to maximize the spectrum efficiency and minimize energy consumption in a wireless ad hoc network. The optimal joint transmission packet scheduling and power control strategy is determined when the node density goes to infinity and the network area is unbounded. Based on the asymptotic analysis, we determine the fundamental capacity limits of a wireless network, subject to an energy consumption constraint. We propose a scheduling and transmission power control mechanism to approach the optimal solution to maximize spectrum and energy efficiencies in a practical network. The distributed implementation of the proposed scheduling and transmission power control scheme is presented based on our medium access control (MAC) framework proposed by Rahimi Malekshan. Simulation results demonstrate that the proposed scheme achieves 40% higher throughput than the existing schemes. In addition, the energy consumption using the proposed scheme is about 20% of the energy consumed using existing power saving MAC protocols.

Efficient radio spectrum utilization and low energy consumption in mobile devices are essential in developing next generation wireless networks. This paper presents a new medium access control (MAC) mechanism to enhance spectrum efficiency and reduce energy consumption in a wireless ad hoc network. A set of coordinator nodes, distributed in the network area, periodically schedule contention-free time slots for all data transmissions/receptions in the network, based on transmission requests from source nodes. Adjacent coordinators exchange scheduling information to effectively increase spatial spectrum reuse and avoid transmission collisions. Moreover, the proposed MAC scheme allows a node to put its radio interface into a sleep mode when it is not transmitting/receiving a packet, in order to reduce energy consumption. Simulation results demonstrate that the proposed scheme achieves substantially higher throughput and has significantly lower energy consumption in comparison with existing schemes.

Energy efficiency is an important performance measure of wireless network protocols, particularly for battery-powered mobile devices such as smartphones. This paper presents a new energy efficient medium access control (MAC) scheme for fully connected wireless ad hoc networks. The proposed scheme reduces energy consumption by putting radio interfaces in the sleep state periodically and by reducing transmission collisions, which results in high throughput and low packet transmission delay. The proposed MAC scheme can also address energy saving in realtime traffics, which require very low packet transmission delay. An analytical model is established to evaluate the performance of the proposed MAC scheme. Analytical and simulation results demonstrate that the proposed scheme has significantly lower power consumption, achieves substantially higher throughput, and has lower packet transmission delay in comparison with existing power saving MAC protocols.

Power consumption is an important consideration in deploying wireless networks, especially for battery-powered mobile devices such as smartphones. Radio interface is the main source of energy consumption in mobile devices. Medium access control (MAC) directly controls the radio interface operation; therefore, an efficient MAC protocol can significantly decrease the energy consumption. This paper presents a new energy efficient MAC scheme for fully connected wireless networks. Time is divided into beacon intervals and a temporary coordinator node manages transmission in each beacon interval. The proposed scheme reduces energy consumption by putting radio interfaces in the sleep state periodically and reducing transmission collision. Simulation results show that the proposed scheme has lower power consumption, higher throughput and lower delay than existing power saving MAC protocols.

In mobile Ad-hoc networks forwarding data through intermediate relays extends the coverage area and enhances the network throughput. Dynamic assignment of routes for data transmission from sources to destinations that is optimized based on channel conditions in a cross-layer manner could provide improved performance. We consider a general multiuser two-hop transmission, where each flow is subject to a constraint on the end-to-end buffering delay and its associated packet loss rate as a quality of service requirement. The objective is to maximize the weighted sum-rate between source destination pairs, while the QoS requirement is satisfied. We introduce two new cross-layer dynamic route selection schemes in this setting that are designed involving physical, MAC, and network layers. In cross-layer opportunistic dynamic route selection, the routes are assigned dynamically based on the instantaneous state of fading channels. In time division cross-layer dynamic route selection, the route for data transmission is scheduled based on the average quality of links. Simulation results show that our proposed cross-layer dynamic route selection schemes provide superior performance in comparison to prior art.