3D Object Detection From Images for Autonomous Driving: A Survey.

3D object detection from images, one of the fundamental and challenging problems in autonomous driving, has received increasing attention from both industry and academia in recent years. Benefiting from the rapid development of deep learning technologies, image-based 3D detection has achieved remarkable progress. Particularly, more than 200 works have studied this problem from 2015 to 2021, encompassing a broad spectrum of theories, algorithms, and applications. However, to date no recent survey exists to collect and organize this knowledge. In this paper, we fill this gap in the literature and provide the first comprehensive survey of this novel and continuously growing research field, summarizing the most commonly used pipelines for image-based 3D detection and deeply analyzing each of their components. Additionally, we also propose two new taxonomies to organize the state-of-the-art methods into different categories, with the intent of providing a more systematic review of existing methods and facilitating fair comparisons with future works. In retrospect of what has been achieved so far, we also analyze the current challenges in the field and discuss future directions for image-based 3D detection research.

Disentangling Monocular 3D Object Detection: From Single to Multi-Class Recognition.

In this paper we introduce a method for multi-class, monocular 3D object detection from a single RGB image, which exploits a novel disentangling transformation and a novel, self-supervised confidence estimation method for predicted 3D bounding boxes. The proposed disentangling transformation isolates the contribution made by different groups of parameters to a given loss, without changing its nature. This brings two advantages: i) it simplifies the training dynamics in the presence of losses with complex interactions of parameters; and ii) it allows us to avoid the issue of balancing independent regression terms. We further apply this disentangling transformation to another novel, signed Intersection-over-Union criterion-driven loss for improving 2D detection results. We also critically review the AP metric used in KITTI3D and resolve a flaw which affected and biased all previously published results on monocular 3D detection. Our improved metric is now used as official KITTI3D metric. We provide extensive experimental evaluations and ablation studies on the KITTI3D and nuScenes datasets, setting new state-of-the-art results. We provide additional results on all the classes of KITTI3D as well as nuScenes datasets to further validate the robustness of our method, demonstrating its ability to generalize for different types of objects.

Reconstruction of the Instantaneous Earth Rotation Vector with Sub-Arcsecond Resolution Using a Large Scale Ring Laser Array.

Absolute rotation rate sensing with extreme sensitivity requires a combination of several large scale gyroscopes in order to obtain the full vector of rotation. We report on the construction and operation of a four-component, tetrahedral laser gyroscope array as large as a five story building and situated in a near surface, underground laboratory. It is demonstrated that reconstruction of the full Earth rotation vector can be achieved with sub-arcsecond resolution over more than six weeks.

Toward the optimization of double-pulse LIBS underwater: effects of experimental parameters on the reproducibility and dynamics of laser-induced cavitation bubble.

Double-pulse laser-induced breakdown spectroscopy (LIBS) was recently proposed for the analysis of underwater samples, since it overcomes the drawbacks of rapid plasma quenching and of large continuum emission, typical of single-pulse ablation. Despite the attractiveness of the method, this approach suffers nevertheless from a poor spectroscopic reproducibility, which is partially due to the scarce reproducibility of the cavitation bubble induced by the first laser pulse, since pressure and dimensions of the bubble strongly affect plasma emission. In this work, we investigated the reproducibility and the dynamics of the cavitation bubble induced on a solid target in water, and how they depend on pulse duration, energy, and wavelength, as well as on target composition. Results are discussed in terms of the effects on the laser ablation process produced by the crater formation and by the interaction of the laser pulse with floating particles and gas bubbles. This work, preliminary to the optimization of the spectroscopic signal, provides an insight of the phenomena occurring during laser ablation in water, together with useful information for the choice of the laser source to be used in the apparatus.

Identification of 14 rare marker chromosomes and derivatives by spectral karyotyping in prenatal and postnatal diagnosis.

Extra structurally abnormal chromosomes (ESACs) and cryptic rearrangements are often associated with mental retardation and phenotypic abnormalities. In some cases their characterisation, using standard cytogenetic techniques and fluorescence in situ hybridization (FISH), is difficult and time consuming, where a fast and accurate identification is essential, especially when such chromosomal aberrations are found in prenatal diagnosis. A recent molecular technique, spectral karyotyping (SKY), based on the spectral signature of 24 chromosome-specific painting probes labelled with different combinations of five fluorochromes, allows the simultaneous visualisation of all human chromosomes in different colours. We used SKY analysis on 14 cases with rare ESACs or cryptic unbalanced rearrangements found at pre- or postnatal diagnosis. SKY analysis permitted the classification of chromosome rearrangements in all 14 cases analysed in combination with FISH analysis.