Search

3D detection of traffic management objects, such as traffic lights and road signs, is vital for self-driving cars, particularly for address-to-address navigation where vehicles encounter numerous intersections with these static objects. This paper introduces a novel method for automatically generating accurate and temporally consistent 3D bounding box annotations for traffic lights and signs, effective up to a range of 200 meters. These annotations are suitable for training real-time models used in self-driving cars, which need a large amount of training data. The proposed method relies only on RGB images with 2D bounding boxes of traffic management objects, which can be automatically obtained using an off-the-shelf image-space detector neural network, along with GNSS/INS data, eliminating the need for LiDAR point cloud data., Comment: Accepted at the 2nd Workshop on Vision-Centric Autonomous Driving (VCAD) as part of ECCV 2024

Autonomous driving is a popular research area within the computer vision research community. Since autonomous vehicles are highly safety-critical, ensuring robustness is essential for real-world deployment. While several public multimodal datasets are accessible, they mainly comprise two sensor modalities (camera, LiDAR) which are not well suited for adverse weather. In addition, they lack far-range annotations, making it harder to train neural networks that are the base of a highway assistant function of an autonomous vehicle. Therefore, we introduce a multimodal dataset for robust autonomous driving with long-range perception. The dataset consists of 176 scenes with synchronized and calibrated LiDAR, camera, and radar sensors covering a 360-degree field of view. The collected data was captured in highway, urban, and suburban areas during daytime, night, and rain and is annotated with 3D bounding boxes with consistent identifiers across frames. Furthermore, we trained unimodal and multimodal baseline models for 3D object detection. Data are available at \url{https://github.com/aimotive/aimotive_dataset}., Comment: The paper was accepted to ICLR 2023 Workshop Scene Representations for Autonomous Driving

Recent studies have demonstrated the efficiency of Variational Autoencoders (VAE) to compress high-dimensional implied volatility surfaces into a low dimensional representation. Although this method can be effectively used for pricing vanilla options, it does not provide any explicit information about the dynamics of the underlying asset. In our work we present an effective way to overcome this problem. We use a Weighted Monte Carlo approach to first generate paths from a simple a priori Brownian dynamics, and then calculate path weights to price options correctly. We develop and successfully train a neural network that is able to assign these weights directly from the latent space. Combining the encoder network of the VAE and this new "weight assigner" module, we are able to build a dynamic pricing framework which cleanses the volatility surface from irrelevant noise fluctuations, and then can price not just vanillas, but also exotic options on this idealized vol surface. This pricing method can provide relative value signals for option traders.

Three-dimensional wide-field galaxy surveys are fundamental for cosmological studies. For higher redshifts (z > 1.0), where galaxies are too faint, quasars still trace the large-scale structure of the Universe. Since available telescope time limits spectroscopic surveys, photometric methods are efficient for estimating redshifts for many quasars. Recently, machine learning methods are increasingly successful for quasar photometric redshifts, however, they hinge on the distribution of the training set. Therefore a rigorous estimation of reliability is critical. We extracted optical and infrared photometric data from the cross-matched catalogue of the WISE All-Sky and PS1 3$\pi$ DR2 sky surveys. We trained an XGBoost regressor and an artificial neural network on the relation between color indices and spectroscopic redshift. We approximated the effective training set coverage with the K nearest neighbors algorithm. We estimated reliable photometric redshifts of 2,879,298 quasars which overlap with the training set in feature space. We validated the derived redshifts with an independent, clustering-based redshift estimation technique. The final catalog is publicly available.

We are presenting a novel, Deep Learning based approach to estimate the normalized broadband spectral energy distribution (SED) of different stellar populations in synthetic galaxies. In contrast to the non-parametric multiband source separation algorithm, SCARLET - where the SED and morphology are simultaneously fitted - in our study we provide a morphology-independent, statistical determination of the SEDs, where we only use the color distribution of the galaxy. We developed a neural network (sedNN) that accurately predicts the SEDs of the old, red and young, blue stellar populations of realistic synthetic galaxies from the color distribution of the galaxy-related pixels in simulated broadband images. We trained and tested the network on a subset of the recently published CosmoDC2 simulated galaxy catalog containing about 3,600 galaxies. The model performance was compared to the results of SCARLET, where we found that sedNN can predict the SEDs with 4-5% accuracy on average, which is about two times better than applying SCARLET. We also investigated the effect of this improvement on the flux determination accuracy of the bulge and disk. We found that using more accurate SEDs decreases the error in the flux determination of the components by approximately 30%., Comment: 5 pages, 8 figures