Nemotron 3 Nano Omni: Efficient and Open Multimodal Intelligence

We introduce Nemotron 3 Nano Omni, the latest model in the Nemotron multimodal series and the first to natively support audio inputs alongside text, images, and video. Nemotron 3 Nano Omni delivers consistent accuracy improvements over its predecessor, Nemotron Nano V2 VL, across all modalities, enabled by advances in architecture, training data and recipes. In particular, Nemotron 3 delivers leading results in real-world document understanding, long audio-video comprehension, and agentic computer use. Built on the highly efficient Nemotron 3 Nano 30B-A3B backbone, Nemotron 3 Nano Omni further incorporates innovative multimodal token-reduction techniques to deliver substantially lower inference latency and higher throughput than other models of similar size. We are releasing model checkpoints in BF16, FP8, and FP4 formats, along with portions of the training data and codebase to facilitate further research and development.

MultiVehicle Simulator (MVSim): lightweight dynamics simulator for multiagents and mobile robotics research

Development of applications related to closed-loop control requires either testing on the field or on a realistic simulator, with the latter being more convenient, inexpensive, safe, and leading to shorter development cycles. To address that need, the present work introduces MVSim, a simulator for multiple vehicles or robots capable of running dozens of agents in simple scenarios, or a handful of them in complex scenarios. MVSim employs realistic physics-grounded friction models for tire-ground interaction, and aims at accurate and GPU-accelerated simulation of most common modern sensors employed in mobile robotics and autonomous vehicle research, such as depth and RGB cameras, or 2D and 3D LiDAR scanners. All depth-related sensors are able to accurately measure distances to 3D models provided by the user to define custom world elements. Efficient simulation is achieved by means of focusing on ground vehicles, which allows the use of a simplified 2D physics engine for body collisions while solving wheel-ground interaction forces separately. The core parts of the system are written in C++ for maximum efficiency, while Python, ROS 1, and ROS 2 wrappers are also offered for easy integration into user systems. A custom publish/subscribe protocol based on ZeroMQ (ZMQ) is defined to allow for multiprocess applications to access or modify a running simulation. This simulator enables and makes easier to do research and development on vehicular dynamics, autonomous navigation algorithms, and simultaneous localization and mapping (SLAM) methods.

Deep Learning Approach to Diabetic Retinopathy Detection

Diabetic retinopathy is one of the most threatening complications of diabetes that leads to permanent blindness if left untreated. One of the essential challenges is early detection, which is very important for treatment success. Unfortunately, the exact identification of the diabetic retinopathy stage is notoriously tricky and requires expert human interpretation of fundus images. Simplification of the detection step is crucial and can help millions of people. Convolutional neural networks (CNN) have been successfully applied in many adjacent subjects, and for diagnosis of diabetic retinopathy itself. However, the high cost of big labeled datasets, as well as inconsistency between different doctors, impede the performance of these methods. In this paper, we propose an automatic deep-learning-based method for stage detection of diabetic retinopathy by single photography of the human fundus. Additionally, we propose the multistage approach to transfer learning, which makes use of similar datasets with different labeling. The presented method can be used as a screening method for early detection of diabetic retinopathy with sensitivity and specificity of 0.99 and is ranked 54 of 2943 competing methods (quadratic weighted kappa score of 0.925466) on APTOS 2019 Blindness Detection Dataset (13000 images).