Back to Courses
Advanced 6 weeks 3 sessions/week

Computer Vision for Production

Build production-grade computer vision systems. From image classification to real-time object detection with YOLO, semantic segmentation, edge deployment, and video analysis pipelines.

£3,497 per person

What You'll Learn

01

Image Classification

Production image classification systems. Modern architectures (EfficientNet, ConvNeXt, Vision Transformers), transfer learning strategies, data augmentation, and handling imbalanced datasets.

02

Object Detection (YOLO, Faster R-CNN)

Real-time object detection with YOLO (v8/v9/v10) and Faster R-CNN. Anchor-based vs. anchor-free detectors, NMS, model selection, and training custom detectors on your own data.

03

Semantic Segmentation

Pixel-level classification with U-Net, DeepLab, and Segment Anything (SAM). Instance segmentation, panoptic segmentation, and building segmentation pipelines for real-world applications.

04

Video Analysis

Process video streams in real-time. Object tracking (SORT, DeepSORT, ByteTrack), action recognition, temporal analysis, and building video analytics pipelines that scale.

05

Edge Deployment (ONNX, TensorRT)

Deploy CV models on edge devices. ONNX export, TensorRT optimisation, quantisation, and running inference on NVIDIA Jetson, Intel OpenVINO, and mobile devices.

06

Real-Time Inference

Build low-latency inference systems. Batching strategies, GPU memory management, model serving with Triton, concurrent request handling, and meeting SLA requirements.

07

Data Augmentation Strategies

Advanced augmentation: Albumentations, CutMix, MixUp, mosaic augmentation, synthetic data generation with diffusion models, and building augmentation pipelines that actually improve performance.

08

Production Pipelines

End-to-end CV pipelines: data labelling workflows, active learning, continuous training, model versioning, A/B testing, and monitoring model quality in production.

Who Is This For

ML engineers who want to specialise in computer vision and deploy CV models in production

Software engineers building products that require image or video analysis capabilities

Robotics and IoT engineers who need to deploy vision models on edge devices

Prerequisites

Python and basic deep learning knowledge required. You should be comfortable with PyTorch, understand CNNs, and have trained neural networks before. Our Deep Learning Masterclass provides ideal preparation.

Course Format

Live Online Sessions

Interactive sessions with real-time Q&A and screen sharing

Recorded Replays

All sessions recorded and available for 12 months after the course

Hands-on Projects

Real-world projects that build your portfolio as you learn

1-on-1 Mentoring

Personal mentoring sessions to address your specific questions

Certificate of Completion

Industry-recognised certificate upon successful completion

Schedule & Pricing

£3,497
6 weeks · 3 sessions per week · 18 sessions total
  • Live interactive sessions
  • 12-month replay access
  • 1-on-1 mentoring
  • Certificate included
Enrol Now

Your Instructors

PP

PeusoPeupon Expert Team

Our instructors are seasoned practitioners with years of experience building production AI systems. They hold certifications across major cloud platforms and have trained thousands of professionals worldwide.

Frequently Asked Questions

GPU access is provided through cloud platforms for all lab sessions. For edge deployment labs, we provide remote access to NVIDIA Jetson devices. A standard laptop is sufficient for the course.

You will build a real-time object detection system, a video analytics pipeline, and deploy a model to an edge device. Each project is designed to mirror real industry applications.

Yes. The techniques (classification, segmentation, detection) are domain-agnostic. We cover domain adaptation and provide guidance for specialised applications including medical and geospatial imaging.

Pre-built APIs are great for common tasks but fall short for custom needs. This course teaches you to build, train, and deploy custom models -- giving you flexibility, better accuracy on your data, and lower per-inference costs at scale.