While recent advancements in camera-based 3D object detection demonstrate remarkable performance, they require thousands or even millions of human-annotated frames. This requirement significantly inhibits their deployment in various locations and sensor configurations. To address this gap, we propose a performant semi-supervised framework that leverages unlabeled RGB-only driving sequences - data easily collected with cost-effective RGB cameras - to significantly improve temporal, camera-only 3D detectors. We observe that the standard semi-supervised pseudo-labeling paradigm underperforms in this temporal, camera-only setting due to poor 3D localization of pseudo-labels. To address this, we train a single 3D detector to handle RGB sequences both forward and backward in time, then ensemble both its forwards and backwards pseudo-labels for semi-supervised learning. We further improve the pseudo-label quality by leveraging 3D object tracking to infill missing detections and by eschewing simple confidence thresholding in favor of using the auxiliary 2D detection head to filter 3D predictions. Finally, to enable the backbone to learn directly from the unlabeled data itself, we introduce an object-query conditioned masked reconstruction objective. Our framework demonstrates remarkable performance improvement on large-scale autonomous driving datasets nuScenes and nuPlan.

We tackle the task of Azimuth (angular dimension) super-resolution for Frequency Modulated Continuous Wave (FMCW) multiple-input multiple-output (MIMO) radar. FMCW MIMO radar is widely used in autonomous driving alongside Lidar and RGB cameras. However, compared to Lidar, MIMO radar is usually of low resolution due to hardware size restrictions. For example, achieving 1° azimuth resolution requires at least 100 receivers, but a single MIMO device usually supports at most 12 receivers. Having limitations on the number of receivers is problematic since a high-resolution measurement of azimuth angle is essential for estimating the location and velocity of objects. To improve the azimuth resolution of MIMO radar, we propose a light, yet efficient, Analog-to-Digital super-resolution model (ADC-SR) that predicts or hallucinates additional radar signals using signals from only a few receivers. Compared with the baseline models that are applied to processed radar Range-Azimuth-Doppler (RAD) maps, we show that our ADC-SR method that processes raw ADC signals achieves comparable performance with 98% (50 times) fewer parameters. We also propose a hybrid super-resolution model (Hybrid-SR) combining our ADC-SR with a standard RAD super-resolution model, and show that performance can be improved by a large margin. Experiments on our Pitt-Radar dataset and the RADIal dataset validate the importance of leveraging raw radar ADC signals. To assess the value of our super-resolution model for autonomous driving, we also perform object detection on the results of our super-resolution model and find that our super-resolution model improves detection performance by around 4% in mAP. The Pitt-Radar and the code will be released at the link.

This work presents a novel dithering algorithm for one dimensional signals. Standard dithering with a triangular probability density white noise increases the noise level by 6dB. This technique is numerically based instead of statistically based, and the increase in noise level can be specified as low as 30. The problem is reduced to a set of non-linear simultaneous equations that are solved using any number of optimization techniques. In this paper, a solution using Levenberg Marquardt is shown and is compared to standard dithering with both real and synthetic signals.

The work presented here deals with the design of predictors for the lossless compression of hyperspectral imagery. The large number of spectral bands that characterize hyperspectral imagery give it properties that can be exploited when performing compression. Specifically, in addition to the spatial correlation which is similar to all images, the large number of spectral bands means a high spectral correlation also. Lossless compression algorithms are typically divided into two stages, a decorrelation stage and a coding stage. This work deals with the design of predictors for the decorrelation stage which are both fast and good. Fast implies low complexity, which was achieved by having predictors with no multiplications, only comparisons and additions. Good means predictors that have performance close to the state of the art. To achieve this, both spectral and spatial correlations are used for the predictor. The performance of the developed predictors are compared to those in the most widely known algorithms, LOCO-I, used in JPEG-Lossless, and CALIC-Extended, the original version of which had the best compression performance of all the algorithms submitted to the JPEG-LS committee. The developed algorithms are shown to be much less complex than CALIC-Extended with better compression performance.