Prerequisites Basic computer literacy and a willingness to document experiments.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Pixels, channels, bit depth and colour spaces Classification, detection, segmentation and keypoints Image provenance, annotation and train/test leakage Resolution, aspect ratio and domain shift Computer Vision specialization: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Tools Python OpenCV Pillow Jupyter
Hands-on exercise Inspect a small image dataset and document class balance, duplicates and acquisition bias. Context for this route: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Suggested project A dataset card and visual audit notebook for a bounded vision task. The outcome must directly demonstrate Computer Vision, not merely list tools.
Learning resources Completion criteria Data sources, labels, resolution and leakage risks are documented. Evidence must be reproducible and tied to Computer Vision.
Next: Image processing and augmentation
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Resize, crop, padding and normalization Colour transforms and geometric operations Augmentation without label corruption Batch loading and reproducibility Computer Vision specialization: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Tools OpenCV Albumentations torchvision
Hands-on exercise Create a preprocessing pipeline and visually compare transformations on representative images. Context for this route: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Suggested project A tested image pipeline with deterministic validation transforms. The outcome must directly demonstrate Computer Vision, not merely list tools.
Learning resources Completion criteria Transformations preserve labels and can be reproduced from configuration. Evidence must be reproducible and tied to Computer Vision.
Next: Classical vision and baselines
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Thresholding, morphology and contours Edges, keypoints and descriptors Image registration and similarity Baseline metrics and error inspection Computer Vision specialization: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Tools OpenCV scikit-image NumPy
Hands-on exercise Build a small inspection task using classical features and evaluate edge cases. Context for this route: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Suggested project A classical-vision prototype with test images and quantitative metrics. The outcome must directly demonstrate Computer Vision, not merely list tools.
Learning resources Completion criteria The baseline has measurable performance and clear limits. Evidence must be reproducible and tied to Computer Vision.
Next: Convolutional networks and transfer learning
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Convolutions, pooling and receptive fields Pretrained backbones and frozen layers Losses, optimizers and regularization Learning curves and class imbalance Computer Vision specialization: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Tools PyTorch torchvision TensorBoard
Hands-on exercise Fine-tune a pretrained classifier and compare it to a simple baseline. Context for this route: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Suggested project A versioned image classifier with confusion matrix and per-class analysis. The outcome must directly demonstrate Computer Vision, not merely list tools.
Learning resources Completion criteria Training is repeatable and the model is evaluated on unseen data. Evidence must be reproducible and tied to Computer Vision.
Next: Object detection and instance segmentation
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Bounding boxes and annotation formats IoU, precision-recall and mAP Detection vs instance segmentation Non-maximum suppression and threshold tuning Computer Vision specialization: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Tools Ultralytics or torchvision detection CVAT or Label Studio COCO evaluation tools
Hands-on exercise Train or fine-tune a detector on a modest labelled dataset and inspect false positives. Context for this route: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Suggested project A documented detection model with sample predictions and mAP evaluation. The outcome must directly demonstrate Computer Vision, not merely list tools.
Learning resources Completion criteria Annotation quality, evaluation protocol and deployment thresholds are documented. Evidence must be reproducible and tied to Computer Vision.
Next: Advanced vision: video, tracking and multimodal
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Video frame sampling and temporal consistency Object tracking and optical flow Pose estimation or OCR as task-specific modules Edge inference and quantization trade-offs Computer Vision specialization: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Tools OpenCV ONNX Runtime MediaPipe where relevant
Hands-on exercise Create a short local video analysis pipeline and measure throughput and stability. Context for this route: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Suggested project A reproducible video or edge-vision prototype with performance measurements. The outcome must directly demonstrate Computer Vision, not merely list tools.
Learning resources Completion criteria The system reports frame rate, failure cases and hardware assumptions. Evidence must be reproducible and tied to Computer Vision.
Next: Robustness, explainability and deployment
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Lighting, blur, occlusion and camera-domain shifts Calibration and confidence thresholds Model export, batching and inference profiling Privacy and responsible image handling Computer Vision specialization: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Tools Captum or saliency tools ONNX Docker
Hands-on exercise Run stress tests on transformed images and document when the model should abstain. Context for this route: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Suggested project A deployment package plus robustness report and model card. The outcome must directly demonstrate Computer Vision, not merely list tools.
Learning resources Completion criteria Limitations and human-review triggers are explicit and measurable. Evidence must be reproducible and tied to Computer Vision.
Next: Capstone: end-to-end vision system
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Dataset specification and annotation QA Training and model selection Inference interface and telemetry Security, privacy and maintenance plan Computer Vision specialization: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Tools PyTorch OpenCV FastAPI Git
Hands-on exercise Build a bounded use-case demo with an auditable dataset split and an inference API. Context for this route: Computer Vision: focus on image pipelines, augmentation, classification, detection/segmentation and spatial evaluation.
Suggested project A complete vision application with reproducible training, evaluation, tests and documentation. The outcome must directly demonstrate Computer Vision, not merely list tools.
Learning resources Completion criteria A reviewer can reproduce the experiment and validate the result on held-out examples. Evidence must be reproducible and tied to Computer Vision.
End of roadmap: Complete the final project and build your portfolio.
✓ Mark this node as complete