Vinitex Laboratory Systems (Netherlands)

Options for treatment of diabetes mellitus in cats are limited to insulin injections and monitoring for hypoglycemia. Once daily sodium-glucose cotransporter-2 inhibitor velagliflozin PO is noninferior to insulin injections. Client-owned diabetic cats (127 safety; 116 efficacy assessment). Prospective, randomized (1 mg/kg velagliflozin), positive controlled (titrated Caninsulin), open label, noninferiority field trial, comparing number of cats with treatment success in ≥1 clinical variable and ≥1 glycemic variable (margin Δ: 15%) on Day 45; secondary endpoints included glycemic and clinical assessments during 91 days. On Day 45, 29/54 (54%) velagliflozin-treated cats and 26/62 (42%) Caninsulin-treated cats showed treatment success, demonstrating noninferiority (difference -11.8%; upper 1-sided 97.5% confidence interval, -∞ to 6.3%). By Day 91, quality of life (QoL), polyuria, and polydipsia had improved in 81%, 54% and 61% (velagliflozin); on blood glucose (BG) curves, mean BG was <252 mg/dL in 42/54 (78%; velagliflozin) and 37/62 (60%; Caninsulin); minimum BG was <162 mg/dL in 41/54 (76%; velagliflozin) and 41/62 (66%; Caninsulin); serum fructosamine was <450 μmol/L in 41/54 (76%; velagliflozin) and 38/62 (61%; Caninsulin). Velagliflozin's most frequent adverse events were loose feces/diarrhea (n = 23/61, 38%), positive urine culture (n = 19/61, 31%), and nonclinical hypoglycemia (BG <63 mg/dL; n = 8/61, 13%); Caninsulin's: clinical and nonclinical hypoglycemia (n = 35/66, 53%), positive urine culture (n = 18/66, 27%), and loose feces/diarrhea (n = 10/66, 15%). Diabetic ketoacidosis occurred in 4/61 (7%; velagliflozin) and 0/66 (Caninsulin). Once daily oral administration of velagliflozin was noninferior to insulin injections, showed good QoL and glycemia without clinical hypoglycemia.

Deep learning models have replaced conventional methods for machine learning tasks. Efficient inference on edge devices with limited resources is key for broader deployment. In this work, we focus on the tool selection challenge for inference deployment. We present an extensive evaluation of the inference performance of deep learning software tools using state-of-the-art CNN architectures for multiple hardware platforms. We benchmark these hardware-software pairs for a broad range of network architectures, inference batch sizes, and floating-point precision, focusing on latency and throughput. Our results reveal interesting combinations for optimal tool selection, resulting in different optima when considering minimum latency and maximum throughput.

Detecting changes in the environment with respect to an earlier data acquisition is important for several applications, such as finding Improvised Explosive Devices (IEDs) and reliable detection of traffic participants. We explore and evaluate the benefit of depth sensing in the context of automatic change detection, where we extend an existing monocular system with a second camera in a fixed stereo setup. We propose an alternative frame registration that exploits scene geometry, in particular the ground plane. Additional feature selection techniques are applied to improve feature distribution on the ground plane, thereby increasing the reliability of the registration. Furthermore, change characterization is applied to localized depth maps to distinguish between physical changes and shadows, which is one of the main challenges of a monocular system. An extensive evaluation on real-world acquisitions containing geo-tagged test objects of 10x10x10 cm is presented in the full paper. Preliminary results show that false positives are reduced by more than 30%.

In this paper, we present a robust and accurate water region detection technique developed for supporting ship identification. Due to the varying appearance of water body and frequent intrusion of ships, a region-based recognition is proposed. We segment the image into perceptually meaningful segments and find all water segments using a sampling-based Support Vector Machine (SVM). The algorithm is tested on 6 different port surveillance sequences and achieves a pixel classification recall of 97.5% and precision of 96.4%. We also apply our water region detection to support the task of multiple ship detection. Combined with our cabin detector, it successfully removes 74.6% false detections generated in the cabin detection process. A slight decrease of 5% in the recall value is compensated by a significant improvement of 15% in precision.