DeepEdge

What DeepEdge does

DeepEdge offers an integrated deep learning development platform that enables developers and enterprises to build, train, optimize, and deploy deep learning models and applications to a variety of edge hardware. It is an industry-first low-code, no-code deep learning development platform purpose-built for edge applications. The platform allows training models in the cloud and deploying them on various edge hardware, including Ambarella, Xilinx, Intel, NVIDIA, and ARM. DeepEdge also provides Platform Services, which offer access to in-house deep learning experts, and Falcon Services, which provide specialized engineering and product development expertise.

Products

• DeepEdge PlatformAn integrated deep learning development platform that enables developers and enterprises to build, train, optimize, and deploy deep learning models and applications to a variety of edge hardware. It is a low-code, no-code platform designed for edge applications.
• DeepDataAn advanced dataset management platform for handling images and videos in various formats, applying filters, performing data health checks, tracking progress, version control, and similarity checks. It supports collaboration and export to standard formats.
• DeepLabelAn AI-assist and active-learning based image labeling platform that supports auto annotation, active learning, team collaboration, single and two-stage validation, and export to different formats. It also allows assigning attributes to dataset labels.
• DeepTrainA low-code, no-code, cloud-based deep learning training platform offering multiple task types (Classification, Segmentation, Object Detection), frameworks (Tensorflow, PyTorch), hyperparameter tuning, experiment tracking, and transfer learning. It has both developer and lite versions.
• DeepDeployAn edge-optimized deployment platform for various edge hardware including Ambarella, Xilinx, Intel, NVIDIA, and ARM. It provides SDKs for specific edge hardware and offers both out-of-box and custom model optimization options.
• DeepZooProvides reference models out-of-the-box for use in model training services.
• Platform ServicesServices that enable full leverage of the DeepEdge platform, offering access to in-house deep learning experts or certified external engineers to accelerate product roadmaps. Includes data services (DeepData, DeepLabel), model training (DeepTrain), model optimization (DeepDeploy), and application development.
• Falcon ServicesHighly specialized engineering services that provide support beyond the DeepEdge platform, including sensor integration, device driver development, system-level engineering, middleware integration, HPC code optimization, custom algorithm development, multimedia expertise, and camera engineering.

Key capabilities

• Integrated deep learning development platform
• Low-code, no-code platform
• Cloud-to-Edge deployment
• Rapid iteration between build, train, and deploy cycles
• Optimize model for performance without losing accuracy
• Seamlessly scale from POC to production
• Architected for scale on AWS
• Seamless collaboration between data scientists, developers, data annotators, and business teams
• Flexibility to train custom models (DeepTrain), import pre-trained models, or select from reference models (DeepZoo)
• No software download required, browser-based operation
• Data-driven model updates
• Support for Ambarella, Xilinx, Intel, NVIDIA, ARM, and Qualcomm (under works) hardware
• Complete IP ownership for uploaded data, custom models, and applications
• Advanced dataset management (DeepData)
• AI-assist and active-learning based image labeling (DeepLabel)
• Multiple task types and frameworks for model training (DeepTrain)
• Hyperparameter tuning and experiment tracking (DeepTrain)
• Edge-optimized deployment (DeepDeploy)
• Model optimization for various edge architectures (DeepDeploy)
• Specialized engineering services (Falcon Services)

Use cases

• Building, training, optimizing, and deploying deep learning models and applications to edge hardware
• Quality inspection of agri products
• Plant disease and pest detection
• Plant growth monitoring
• Intelligent analysis for indoor farming
• Livestock management
• Aerial crop monitoring
• Weed detection
• Reducing losses during harvesting (e.g., tomatoes)
• Monitoring workplace safety
• Predictive maintenance for equipment
• Quality control and defect detection in manufacturing (e.g., scratches, dents, weld, paint flaws)
• Assembly monitoring in automotive, electronics, and packaging industries
• Defect identification in production lines (e.g., sheet metal)
• Automating eyelash extension application
• Manicure applications
• Virtual fitting room applications
• Digital make-up applications
• Apparel and accessories recognition in video and images
• Warehouse safety and security (employee access, PPE usage, hazard identification)
• Improving employee productivity in warehouses
• Pallet and forklift monitoring
• Facial recognition on battery-operated body-worn cameras
• Fall detection on battery-operated smart walkers for elderly people
• Retail planograms application

AI approach

DeepEdge provides an integrated deep learning development platform for building, training, optimizing, and deploying deep learning models and applications to various edge hardware. It offers low-code, no-code tools for data management (DeepData), annotation (DeepLabel with AI-assist and active learning), model training (DeepTrain with transfer learning and custom model development), and deployment (DeepDeploy for edge optimization). They also offer Platform Services and Falcon Services, which include deep learning expertise, data services, model training, optimization, and application development, as well as specialized engineering services for sensor integration, system-level engineering, and HPC optimization.

Tech named: Deep Learning, Computer Vision, EdgeAI, Custom CNN model, Tensorflow, PyTorch, Ambarella, Xilinx, Intel, NVIDIA, ARM, Qualcomm, HPC, ROS - Robotics Operating System, PNG, JPEG, HEVC, H264, AVI, MP4, GStreamer

Industries served

• Agriculture
• Manufacturing & Industrial
• Beauty & Fashion
• Warehouse
• Industrial IoT (IIoT)
• Robotics
• Law Enforcement
• Elderly Care
• Retail

What it says sets it apart

• Industry's first low-code, no-code deep learning development platform purpose-built for Edge applications
• One integrated deep learning development platform
• Train in the cloud and deploy on any edge hardware
• Rapid iteration between build, train, and deploy cycles
• Reduces hardware complexity
• Optimizes model performance without losing accuracy
• Architected for scale on AWS
• Seamless collaboration features
• Flexible model training options (custom, import, reference)
• Browser-based, no software download needed
• Data-driven continuous model updates
• Broad hardware support including Ambarella, Xilinx, Intel, NVIDIA, ARM, and Qualcomm
• Complete IP ownership for customers
• AI-assist and active-learning in DeepLabel for cost-quality ratio
• Specialized engineering and product development expertise through Falcon Services