Course Objective & Scope
In production settings with scarce labels, imbalanced classes and changing conditions, the classic supervised approach falls short. This advanced programme covers self-supervised backbones such as DINOv2, foundation models such as SAM 2 and Grounding DINO, open-vocabulary detection and few-shot approaches.
Synthetic data generation, active learning loops, handling domain shift and multi-GPU training (PyTorch DDP) are covered hands-on. Participants prepare a foundation-model-based solution architecture and experiment plan for a scenario close to their own problems.
Who Should Attend?
- Computer vision engineers who have run CNN and detection models in production
- AI specialists in research and development teams
- Technical leads responsible for visual AI architecture
- Prerequisite: experience training models with PyTorch
Key Takeaways
- Evaluate self-supervised backbones on sparsely labelled data
- Cut labelling costs with SAM 2 and Grounding DINO
- Prioritise samples for labelling with active learning
- Detect domain shift and build a mitigation strategy
- Configure and scale multi-GPU training
Curriculum
Self-Supervised Learning and Strong Backbones
- Contrastive and masked image modelling approaches
- Linear probing with DINOv2 features
Vision Foundation Models
- SAM 2, Grounding DINO and open-vocabulary detection
- Zero-shot and few-shot usage scenarios
Data-Centric Approaches
- Data expansion with synthetic data and simulation
- Active learning and automatic pre-labelling
Domain Shift and Robustness
- Robustness to lighting, camera and environment changes
- Test-time augmentation and domain adaptation
Large-Scale Training
- PyTorch DDP and gradient accumulation
- Experiment tracking and reproducibility
Other Courses in This Category
-
Object detection and instance segmentation with YOLO, RT-DETR and Mask R-CNN: labelling, training, mAP-based evaluation and tracking integration, for teams building an end-to-end detection/segmentation prototype.
- Live Course
- 5 modules
-
Computer Vision Bootcamp
Comprehensive Computer Vision Bootcamp ranging from classical image processing techniques to PyTorch-based CNN, Vision Transformer (ViT), Swin Transformer, and ConvNeXt architectures. Hands-on projects cover object detection with YOLO and DETR, segmentation with SAM and U-Net, video analysis, 3D vision (NeRF, Gaussian Splatting), Diffusion models, and model deployment with ONNX/TensorRT.
- Live Course
- 13 modules
-
Semantic, instance and panoptic segmentation: pixel-level training with U-Net, DeepLabV3+, SegFormer, Mask2Former and SAM 2, Dice/mIoU evaluation and mask-based measurement, for industrial, medical and satellite imagery teams.
- Live Course
- 6 modules
-
ViT, DeiT, Swin and DINOv2 architectures: patch embedding, self-attention, comparison with CNNs and fine-tuning with Hugging Face, for engineers selecting and adapting transformer-based models for vision tasks.
- Live Course
- 5 modules
-
Document preprocessing, text recognition, layout analysis and table/form extraction with PaddleOCR, Surya, LayoutLMv3 and Donut, for technical teams turning invoices, contracts and operational documents into structured data.
- Live Course
- 5 modules
-
RTSP video streams, GStreamer/FFmpeg, real-time detection, multi-object tracking with ByteTrack/BoT-SORT, speed/crowd analytics and NVIDIA DeepStream, for teams extracting live operational insights from camera networks.
- Live Course
- 5 modules
Get in touch
