Course Objective & Scope
Traditional computer vision models require thousands of labelled images and only recognise fixed classes, whereas Vision-Language Models (VLMs) can reason zero-shot over complex scenes, compare multiple photos and explain their findings. This course teaches how to build production VLM applications over real-world product and field imagery.
Participants build two-stage pipelines combining fast detectors (YOLO/SAM) with VLMs, apply visual prompting, bounding-box grounding and multi-angle comparison, and deliver a visual analysis application tailored to an organisational use case.
Who Should Attend?
- Computer vision and AI application engineers
- Technology teams in retail, e-commerce, manufacturing, insurance and field operations
- Solution architects leading visual automation projects
- Prerequisite: Python, basic image handling (OpenCV/Pillow) and API familiarity
Key Takeaways
- Combine fast object detectors with VLMs in two-stage pipelines to balance speed, cost and reasoning
- Boost spatial accuracy with visual prompting, cropping and Set-of-Marks overlays
- Build workflows that compare before/after or reference/actual multi-image sets
- Generate schema-validated JSON outputs for catalogue enrichment, shelf auditing or damage assessment
- Implement image quality gates (blur, lighting, framing) and evaluation benchmarks
Curriculum
Synergy Between Classic Computer Vision and VLMs
- When to use YOLO/classifiers, when to use VLMs and when to combine them
- Detect-then-crop architectures that pair fast local models with VLM reasoning
Spatial Grounding and Visual Prompting
- Extracting 2D bounding boxes and point coordinates (Gemini, Qwen-VL, Florence-2)
- Overlaying numbered markers and grids to improve counting and spatial accuracy
Multi-Image Comparison and Video Clip Analysis
- Before/after field verification, planogram compliance and damage comparison
- Keyframe extraction and temporal reasoning across short inspection videos
Image Quality Gates and Robustness
- Pre-checks for blur, poor lighting, wrong angles and recaptured/spoofed photos
- Prompting field users in real time to retake unusable photos
Lab: Enterprise Visual Inspection Prototype
- End-to-end visual analysis application on a catalogue, field or claims case
- IoU, classification F1 and per-image cost/latency measurement plan
Other Courses in This Category
-
End-to-end Speech-to-Speech (S2S) models and autonomous voice agents: acoustic emotion and prosody perception, voice tool calling, multi-agent call handoffs and production architecture with Pipecat and LiveKit.
- Live Course
- 5 modules
-
Streaming infrastructure engineering for real-time AI applications: WebRTC, WebSockets and SSE protocols, OpenAI Realtime and Gemini Live API integration, audio/video buffer management and horizontal scaling.
- Live Course
- 5 modules
-
Human-like turn-taking and millisecond latency engineering in voice AI: acoustic VAD, semantic endpointing, false-interruption suppression, TTFT/TTFB optimisation and speculative execution.
- Live Course
- 5 modules
-
Parsing dense documents and application screens (GUIs) at the pixel, coordinate and structural-tree level: OmniParser, Qwen-VL, Docling, UI widget detection and unified document/screen understanding models.
- Live Course
- 5 modules
-
Evaluation engineering for multimodal systems across vision, documents, audio and video: OCR/extraction metrics, grounding IoU, speech WER/MOS and turn-taking metrics, VLM-as-a-judge and regression pipelines.
- Live Course
- 5 modules
-
Comprehensive introduction to multimodal AI models that understand and generate text, images, audio and video: model families, enterprise use cases, capability limits, cost/latency trade-offs and model selection.
- Live Course
- 5 modules
Get in touch
