AI-Powered Vision Analytics Platform

AI-Powered Vision Analytics Platform

DeepVi is a No-Code, End-to-End AI Vision Platform that handles everything from data collection to model training, precision inspection, deployment, and retraining — all within a single platform. Training and inference are monitored and controlled at a glance through the Built-in Workflow, and workflows are extended and edited with the VCANUS workflow editor. Alongside deep learning (Classification · Detection · Segmentation), DeepVi includes Align · Inspection (alignment, measurement, and code reading) — precision inspection functions widely used in image analysis. No dedicated development team required: field operators can run the full labeling, training, inspection, inference, and review cycle themselves. It supports both on-premises and cloud deployments and provides a universal pipeline ready for immediate use across industries — including manufacturing, smart farming, logistics, and security.

What You Can Do with DeepVi

Application Areas

The same pipeline works across industries — just change the class names and data, and DeepVi extends immediately beyond manufacturing, smart farming, logistics, and security to medical & healthcare, optical sorting, embedded vision, automated test equipment (ATE), and more.

Key Features

Storage (Data Management)

• Bulk web upload for registering large image sets
• Folder-based management — create, move, and delete folders; bulk-move multiple selections
• High-speed browsing of large datasets with auto-generated thumbnails
• Storage usage visible at a glance
• Original data preserved permanently — easy to reuse
• Datasets reference the originals — storage footprint stays constant as datasets are added

Dataset (Labeling)

• Unified labeling environment for all 3 types: Classification · Detection · Segmentation
• BBox · Polygon · Mask precision labeling tools
• AI-assisted segmentation and Auto Labeling to cut labeling time
• Import/export in standard formats such as COCO, plus category templates
• Published dataset snapshots (fixed train/val/test split) for reproducible training

Training · Model (Workflow Training & Validation)

• Built-in Workflow — training flow, node status, and logs on a single screen
• Per-node run · pause · resume, plus parameter adjustment
• Workflows extended and edited with the VCANUS workflow editor
• Fast / Standard model families, with more families addable
• Anomaly Detection — training for structural and logical anomalies
• Transfer Learning and Early Stopping, with a real-time Loss chart
• Performance evaluation with Confusion Matrix and mAP; checkpoint and ONNX export

Align · Inspection (Precision Inspection)

• Reference image registration and one-click Align model creation
• Automatic shift and rotation correction (Align)
• 6 inspection tools: Caliper · Circle · Distance · Angle · Code · Template
• Tolerance-based OK/NG judgment, with a golden dry-run on the reference image
• Real-time judgment alongside deep-learning results in the inference workflow

Inference (Real-Time Inference)

• Live inference from camera streams or image files
• Workflow-based interface for fast algorithm configuration (add custom nodes)
• Real-time display of detections and confidence, plus pre-processing / inference / post-processing time analysis
• Detection, segmentation, and classification results shown together with Align · Inspection OK/NG
• Results, coordinates, and timestamps auto-saved; history lookup with condition filters
• Result-driven actions — trigger equipment control, data logging, alerts and more as workflow actions

Review (Model Improvement)

• Build up training data through false-positive review, re-labeling, and comments
• Review status management — approve, reject (with reason codes), or discard, with per-item history
• Compare and edit reviewed labels against the original predictions
• Reviewed data reflected into the dataset → retraining
• After performance evaluation, the production model is replaced with a verified model

DeepVi in One Line

No coding required, all in one platform — DeepVi removes every barrier to AI adoption.

DeepVi handles the complete End-to-End AI vision pipeline in a single platform: data collection (Storage) → labeling (Dataset) → workflow training (Training) → precision inspection (Align · Inspection) → real-time inference (Inference) → model improvement (Review). The same pipeline works across industries — just change the class names and the data, and it adapts immediately to manufacturing, smart farming, logistics, security, medical imaging, and more.