Superior Industrial Vision Language Model: AQ-VLM
AQ-VLM is a Vision Language Large Model (VLM) dedicated for industrial quality control challenges for small-batch, multi-product production. It enables unified model deployment for classification, detection, segmentation, and OCR, delivering more reliable and scalable solution compared to industrial small models.
Product Highlights
High Accuracy
Precisely captures critical features in industrial images with 99%+ accuracy, providing more reliable quality assurance than small models.
Full Product Compatibility
With strong generalization capabilities, AQ-VLM can cover all sub-variants, eliminating the need for frequent retraining of AI models.
No Training Required
After collecting 2,000-5,000 images per defect type, AQ-VLM can become stable L2 domain-specific large model without further training.
Core Technology
Based on the generated data accumulated in 8+ years, AQ-VLM has become a stable L1 industrial large model and can be fine-tuned with the customers' data to quickly generate the L2 domain-specific VLM that adapts to different customer needs.
AQ-Data
Aggregates data from multiple industrial scenarios, using a data engine to ensure high-quality, large-scale datasets.
AQ-LoRA
Low-rank adaptation (LoRA) technology enables rapid fine-tuning with minimal data.
AQ-Image Encoder
Optimized image encoder for industrial data, extracting fine-grained features for precise defect recognition.
Real-World Applications
Magnet Material Inspection
Covers all sizes, shapes, and coatings. New product variants can be inspected immediately, achieving 0% underkill rate for severe defects and 0.5% underkill rate for minor defects.
Connector Inspection
Supports 100+ connector variants, automatically identifying defects with 99% accuracy, ensuring product consistency and quality compliance.
Battery Electrode Sorting
Handles variations in size, lighting, and texture. Quickly identifies defective electrodes with 0% underkill rate and 1% overkill rate, optimizing recall processes on the production line.
Wine Box SKU Recognition
Despite changes in positions and damages, AQ-VLM automatically recognizes printed text and matches it to its warehouse SKU with 99.9% accuracy, increasing warehouse sorting efficiency by 30%.
Personnel Dress Code Check
Check potential violations for hygiene and safety regulations in factories with 99% accuracy, promptly alarm and improve factory management.
Get Started with AQ-VLM
Step 1
Step 2
Step 3
Step 4
