Localisation for Computer Vision Engineers
Localisation tasks ask not only what is in an image but where. Object detection predicts bounding boxes with class labels, using methods such as Faster R-CNN or YOLO. Segmentation labels pixels, either by class or by individual object. Both rely on a CNN or transformer backbone plus task-specific heads and evaluation metrics.
Stage: Vision Tasks · Part of the Computer Vision Engineer roadmap · Last reviewed 19 November 2026
What to learn in Localisation
- Compute intersection over union between a predicted and a ground-truth box
- Apply non-maximum suppression to remove duplicate detections
- Explain the difference between one-stage and two-stage detectors
- Distinguish semantic, instance and panoptic segmentation
- Evaluate a detector with mean average precision at an IoU threshold
Localisation interview questions
What is intersection over union and how is it used?
IoU is the area where two boxes overlap divided by the area of their union, giving a value from 0 to 1. It measures localisation quality. Detectors use it to match predictions to ground truth, to run non-maximum suppression, and to define true positives in average precision.
What is the difference between semantic and instance segmentation?
Semantic segmentation assigns a class to every pixel but does not separate individual objects, so two adjacent people share one label region. Instance segmentation also distinguishes each object, giving separate masks for each person, as in Mask R-CNN.
Where Localisation fits in the Computer Vision Engineer roadmap
Localisation belongs to the Vision Tasks stage (advanced level), which takes about 38 hours. Beyond classification.
Other units in this stage
Back to the full Computer Vision Engineer roadmap