PLAltform - Cosmo

PLAItform

AI Capabilities Platform

With Orwell underlying algorithm capabilities, WellSpiking one-stop AI development platform,

and WellData integrated data platform, we form a solid AI base to support all Westwell services.

Orwell

WellSpiking

WellData

A smarter future, now

Inspiration for Intelligence of Everything

Starting point for Internet of Everything

Building a strong smart hub for all intelligent possibilities

Support quantification, pruning, sparseness, and training optimization of most custom algorithm models to maximize the model accuracy

Support mainstream open-source deep learning computing frameworks, which allow users to integrate their own frameworks easily

Provides a series of hardware deployment, calling, debugging, and monitoring tools, which helps users develop products and maintain projects

Provides various types of open-source and general task algorithm libraries, which can be trained directly and deployed on hardware platforms

Provides hardware implementation, optimization, export, and simulation tools for a series of algorithm models, which allows users to optimize hardware algorithms independently and effectively isolate service models and data

Orwell VCU V100
Decoding Unit

Dedicated image decoding card. Embedded with a dedicated H.264/H.265 video stream encoding and decoding unit, it supports decoding, image revision, and other operations across 16 video streams to meet different AI algorithms' requirements for input images.

Orwell PAU V200
Acceleration Unit

Dedicated AI acceleration card. Embedded with a HardRock AI acceleration unit, it supports various deep learning operators, flexibly orchestrates deep learning networks, and supports various AI models such as target detection, image recognition, and semantic segmentation. Features a PCIe Gen3x8 interface, has a power consumption of 30 W, and can work in a temperature range of -20°C to +70°C.

Orwell
Edge Acceleration Unit

The Orwell edge inference acceleration unit is a heterogeneous computing system based on Intel CPU and FPGA. It supports real-time encoding, decoding, and inference acceleration across 16 cameras and supports Int8/16 quantified deep learning operations.

Wellseesee
Industrial-Grade Dual-lens Stereoscopic Camera

Meets the requirements of 3D measurement, target identification, distance measurement, and image segmentation and has been verified in Westwell application scenarios.

Millisecond-level imaging

Wide dynamic range

Super strong processing power,

40 m line-of-sight

IP67 rated

Low-light imaging

One-stop AI development platform

Promoter for industry empowerment with AI

Serves port, train, city operation, and other scenarios

Provides massive data preprocessing and marking, large-scale distributed training,

automatic model generation and testing for machine learning and deep learning

Full-Process Management

Integrate data collection and processing, model training, testing, and deployment functions to quickly build all required AI models.

Deep Learning

Support mainstream deep learning frameworks such as TensorFlow, Caffe, and Torch. Also support GPU distributed computing using either one machine and multiple cards or both multiple machines and cards.

Excellent Performance

Embed with a large number of CPU/GPU/FPGA servers with 10G NICs and acceleration algorithms for distributed machine learning to provide a solid foundation for model training.

Simple and Convenient Operations

The flexible and agile command line operation mode adapts to the habits of professional users. The visual operation mode allows drag-and-drop of algorithm components to implement service logic. The interface is user-friendly and easy to use.

Data factor flowing
Intelligent data insight
Data value mining

WellData integrates data from various scenarios, services,

and terminals to ensure real-time, intelligent, and efficient data analysis

Multi-terminal, Multi-scenario Data Aggregation
Providing Comprehensive and Accurate Data Support

Multi-terminal

Data collection from cameras, IoT devices, PLCs, and other channels

Multi-scenario

Data integration in multiple scenarios, such as ports, airports, campuses, and communities

Systematic Data Labels, Supporting In-time Service Query

Based on service models and scenarios, users can attach labels to data to facilitate subsequent data analysis

Unified Data Base, Supporting Agile Upper-Layer Development

The unified data standard ensures orderly data aggregation Standard data can be obtained in an agile manner from the unified resource pool to quickly respond to changeable requirements

Learn more about our smart solutions
Continue to explore「Novus」

Learn More