Deep neural networks have become a promising approach for IMU-based sensing, but their scalability is fundamentally limited by costly labeled data and poor robustness to heterogeneous devices, placements, and users. Existing unsupervised and self-supervised methods reduce but do not remove this dependence, still requiring labeled data for domain adaptation and largely ignoring known physical structure. We propose physical self-supervised learning, an autoencoder-style paradigm for label-free IMU sensing. We replace the conventional neural decoder with an auto-adaptive physics decoder—a learnable family of kinematic equations that enforces explicit physical structure while adapting across environments—and adopt a hybrid two-stage IMU encoder with reconstruction in a structured latent space to mitigate sensor noise. Our framework further introduces probabilistic frequency-spatial constraints to disentangle sensor and object motion, a multi-view kinematic tree to exploit sparse physical self-supervised signals, and an uncertainty-aware formulation to handle the inherent ambiguity of IMU inference. Evaluated on inertial tracking and full-body motion capture over public datasets and realistic deployments, physical self-supervised learning reduces errors by up to 5× for tracking and 4× for motion capture in challenging generalization scenarios, consistently outperforming state-of-the-art supervised and self-supervised baselines without any labels.

Recent advancements in on-device training for deep neural networks have underscored the critical need for efficient activation compression to overcome the memory constraints of mobile and edge devices. As activations dominate memory usage during training and are essential for gradient computation, compressing them without compromising accuracy remains a key research challenge. While existing methods for dynamic activation quantization promise theoretical memory savings, their practical deployment is impeded by system-level challenges such as computational overhead and memory fragmentation.

Recent advancements in exploring machine learning models' dynamic spatial sparsity have demonstrated great potential for superior efficiency and adaptability without compromising accuracy when compared to conventional static-and-dense DNNs. However, realizing theoretical inference acceleration under practical deployment environments is still faced with significant system challenges. Current vendor libraries and tensor compilers fall short due to their extra data copy operations or insufficient computation schemes, especially for DNN operators with dynamic spatial sparsity.

Deep visual data processing is underpinning many life-changing applications, such as auto-driving and smart cities. Improving the accuracy while minimizing their inference time under constrained resources has been the primary pursuit for their practical adoptions. Existing research thus has been devoted to either narrowing down the area of interest for the detection or miniaturizing the deep learning model for faster inference time. However, the former may risk missing/delaying small but important object detection, potentially leading to disastrous consequences (e.g., car accidents), while the latter often compromises the accuracy without fully utilizing intrinsic semantic information. To overcome these limitations, in this work, we propose ScaleFlow, a closed-loop scale-adaptive inference that can reduce model inference time by progressively processing vision data with increasing resolution but decreasing spatial size, achieving speedup without compromising accuracy. For this purpose, ScaleFlow refactors existing neural networks to be scale-equivariant on multiresolution data with the assistance of wavelet theory, providing predictable feature patterns on different data resolutions. Comprehensive experiments have been conducted to evaluate ScaleFlow. The results show that ScaleFlow can support anytime inference, consistently provide 1.5× to 2.2× speed up, and save around 25% ~ 45% energy consumption with < 1% accuracy loss on four embedded and edge platforms

Smart cities, utilities, third-parties, and government agencies are having pressure on managing stochastic power generation from distributed rooftop solar photovoltaic (PV) arrays, such as predicting and reacting to the variations in electric grid. Recently, there is a rising interest to identify solar PV arrays automatically and passively. Traditional approaches such as online assessment and utilities interconnection filings are time consuming and costly, and limited in geospatial resolution, and thus do not scale up to every location. Significant recent work focuses on using aerial imagery to train machine learning or deep learning models to automatically detect solar PV arrays. Unfortunately, these approaches typically require Very High Resolution (VHR) images and human handcrafted solar PV array templates for training, which have a minimum cost of $15 per km2and are not always available at every location.To address the problem, we design a new system—SolarFinder that can automatically detect distributed solar PV arrays in a given geospatial region without any extra cost. SolarFinder first automatically fetches regular resolution satellite images within the region using publicly-available imagery APIs. Then, SolarFinder leverages multi-dimensional K-means algorithm to automatically segment solar arrays on rooftop images. Eventually, SolarFinder employs hybrid linear regression approach that integrates support vector machine (SVM) modeling with a deep convolutional neural networks (CNNs) approach to accurately identify solar PV arrays and characterize each solar deployment. We evaluate SolarFinder using 269,632 satellite images that include 1,143,636 contours from 13 geospatial regions in U.S. We find that pre-trained SolarFinder yields a Matthews Correlation Coefficient (MCC) of 0.17, which is 3 times better than the most recent pre-trained CNNs approach and the same as a re-trained CNNs approach.