Absolute Distance Prediction Based on Deep Learning Object Detection and Monocular Depth Estimation Models

Determining the distance between the objects in a scene and the camera sensor from 2D images is feasible by estimating depth images using stereo cameras or 3D cameras. The outcome of depth estimation is relative distances that can be used to calculate absolute distances to be applicable in reality. However, distance estimation is very challenging using 2D monocular cameras. This paper presents a deep learning framework that consists of two deep networks for depth estimation and object detection using a single image. Firstly, objects in the scene are detected and localized using the You Only Look Once (YOLOv5) network. In parallel, the estimated depth image is computed using a deep autoencoder network to detect the relative distances. The proposed object detection based YOLO was trained using a supervised learning technique, in turn, the network of depth estimation was self-supervised training. The presented distance estimation framework was evaluated on real images of outdoor scenes. The achieved results show that the proposed framework is promising and it yields an accuracy of 96% with RMSE of 0.203 of the correct absolute distance.

A differential RR lyrae line-of-sight distance between M31 and M33

We present a purely differential line-of-sight distance between M31 and M33 using ab-type RR Lyrae variables observed in each galaxy by the Hubble Space Telescope Advanced Camera for Surveys in the F606W filter. Using 1481 RR Lyraes in 13 M31 fields and 181 RR Lyraes in 6 M33 fields, and placing all of these stars on a uniform photometric scale with internally consistent corrections for metal abundance and extinction, we find a relative absolute distance modulus of Δ(m-M)o = –0.298 ± 0.016 in the sense of (m-M)o, M31 – (m-M)o, M33. Adopting an absolute distance modulus of (m-M)o=24.46±0.10 for M31 places M33 115 kpc beyond M31 in line-of-sight distance.

Quantitative Motion Measurements Based on Markerless 3D Full-Body Tracking in Children with SMA Highly Correlate with Standardized Motor Assessments

Background: Spinal Muscular Atrophy (SMA) is the most common neurodegenerative disease in childhood. New therapeutic interventions have been developed to interrupt rapid motor deterioration. The current standard of clinical evaluation for severely weak infants is the Children’s Hospital of Philadelphia Infant Test of Neuromuscular Disorders (CHOP INTEND), originally developed for SMA type 1. This test however, remains subjective and requires extensive training to be performed reliably. Objective: Proof of principle of the motion tracking method for capturing complex movement patterns in ten children with SMA. Methods: We have developed a system for tracking full-body motion in infants (KineMAT) using a commercially available, low-cost RGB-depth sensor. Ten patients with SMA (2–46 months of age; CHOP INTEND score 10–50) were recorded for 2 minutes during unperturbed spontaneous whole-body activity. Five predefined motion parameters representing 56 degrees of freedom of upper, lower extremities and trunk joints were correlated with CHOP INTEND scores using Pearson product momentum correlation (r). Test-retest analysis in two patients used descriptive statistics. Results: 4/5 preselected motion parameters highly correlated with CHOP INTEND: 1. Standard deviation of joint angles (r = 0.959, test-retest range 1.3–1.9%), 2. Standard deviation of joint position (r = 0.933, test-retest range 2.9%), 3. Absolute distance of hand/foot travelled (r = 0.937, test-retest range 6–10.5%), 4. Absolute distance of hand/foot travelled against gravity (r = 0.923; test-retest range 4.8–8.5%). Conclusions: Markerless whole-body motion capture using the KineMAT proved to objectively capture motor performance in infants and children with SMA across different severity and ages.