AI powered LiDAR analytics detect, classify, and track traffic movement in real time, providing accurate insights into vehicle and pedestrian flow. The solution enables smarter traffic management and supports future ready urban planning.
VeSoniq is Vedya Labs’ Voice UI solution running on the Microchip SAMA7D65 platform. It enables always-on keyword spotting, custom wake words, and real-time voice command processing, delivering fast and reliable voice interaction for embedded and Edge AI applications.
VeSoniq is Vedya Labs’ Voice UI solution designed to run across multiple embedded platforms. This demo showcases always-on keyword spotting and real-time voice command processing on the Microchip dsPIC33A, delivering responsive on-device voice interaction for embedded and Edge AI applications.
Vedya Labs demonstrates Stable Diffusion text-to-image generation on the Axera AX650N SoC using dual Cadence Tensilica Vision Q7 DSP cores. Optimized for DSP-centric execution with INT8 quantization and efficient memory scheduling, the solution enables GPU/NPU-independent Generative AI inference at the edge.
An autonomous delivery bot built on the Renesas RZ platform and powered by VeRNOX (Robotics SDK from Vedya Labs), combining perception, Visual SLAM, navigation, motion planning, and real-time motor control to autonomously pick up and deliver objects.
Vedya Face ID is an edge-native face recognition solution built with lightweight INT8 AI models for embedded deployment. Using BlazeFace for face detection and Shuffle FaceNet for recognition, it achieves <200 ms end-to-end latency, >95% recognition accuracy, and <1% false positive rate, with robustness across low-light and varied appearance conditions.
VeSoniq, Vedya Labs’ Voice UI solution, runs on the Renesas RA8P1 with a lightweight, fully quantized model under 150 KB. With noise resilience down to −10 dB SNR, up to 3 wake words and 30 commands, few-shot learning, local language support, and speaker-aware personalization, it enables responsive on-device voice interaction.
Vedya Labs deploys SmolVLA (~450M parameters) on NVIDIA Jetson AGX Orin 64 GB to control an SO-101 robotic arm. Using three camera views, the system picks up a red eraser and places it inside a target, demonstrating edge AI inference and vision-language-action-based robotic manipulation.
