Watch for Camera Drift
A camera slowly tilts after maintenance.
Read the full lesson and open its transcript below. A narrated video will be added after production.
What you will learn
A camera slowly tilts after maintenance.
A practical question
Months after installation, a parking camera slowly tilts. Its predicted bay polygons still appear on screen, but their edges drift onto the driving lane. That is a physical change in the input, not merely a model score change.
The useful idea
Monitor stable landmarks, image framing, exposure, and output patterns over time. Sudden changes can follow maintenance; gradual shifts can follow a loose mount. A surge in uncertain cases or corrections can be another clue, but it does not identify the cause on its own.
Work through the example
Compare a recent frame with the original calibrated view. Are painted corners still in the same positions? If not, recalibrate or pause automatic decisions that depend on those polygons. Keep examples and timestamps of the shift. A camera that slowly shifts can move the drawn parking regions away from the actual lines. Track stable landmarks and sample a few human-labeled frames over time. If both landmark positions and occupancy errors drift, investigate mounting or calibration before retraining the detector. Record when the shift began so affected counts can be identified. A monitoring chart should prompt a specific inspection, not automatically change a working model. Use this case to test your understanding: Compare stable landmark positions and sampled human labels over time. Here is the limitation to watch for: A stable average occupancy count could still hide a shifted camera. The answer follows from that evidence: Drift monitoring tracks changes in viewpoints, lighting, data, and error rates. Compare landmark positions before and after the camera tilt. A changed space map may look like a detector failure when the underlying boxes are sound. Keep timestamped frames so affected counts can be identified. Compare with a reference view taken before maintenance to locate the shift.
Limits and next step
Next, we'll plan how to update the vision system without letting a well-intended fix quietly introduce new errors.
Worked example
Compare stable landmark positions and sampled human labels over time.
Quick check: test your understanding
Try it yourself: Consider this case: Compare stable landmark positions and sampled human labels over time. Explain how you would check the result. In your answer, address this limitation: A stable average occupancy count could still hide a shifted camera.
Check the reasoning
Drift monitoring tracks changes in viewpoints, lighting, data, and error rates. A stable average occupancy count could still hide a shifted camera.
What is the main idea behind watch for camera drift?
Drift monitoring tracks changes in viewpoints, lighting, data, and error rates.
How is it used in the example?
Compare stable landmark positions and sampled human labels over time.
What limitation should be checked?
A stable average occupancy count could still hide a shifted camera.
Further reading
Programming Computer Vision with Python — Jan Erik Solem.
Original ToolsInVision teaching examples. Editorial approach · Suggest a correction
Accessible referenceLesson 75 transcript302 words · Narration ready for the upcoming videoOpen transcript
Months after installation, a parking camera slowly tilts. Its predicted bay polygons still appear on screen, but their edges drift onto the driving lane. That is a physical change in the input, not merely a model score change.
Monitor stable landmarks, image framing, exposure, and output patterns over time. Sudden changes can follow maintenance; gradual shifts can follow a loose mount. A surge in uncertain cases or corrections can be another clue, but it does not identify the cause on its own.
Compare a recent frame with the original calibrated view. Are painted corners still in the same positions? If not, recalibrate or pause automatic decisions that depend on those polygons. Keep examples and timestamps of the shift. A camera that slowly shifts can move the drawn parking regions away from the actual lines. Track stable landmarks and sample a few human-labeled frames over time. If both landmark positions and occupancy errors drift, investigate mounting or calibration before retraining the detector. Record when the shift began so affected counts can be identified. A monitoring chart should prompt a specific inspection, not automatically change a working model. Use this case to test your understanding: Compare stable landmark positions and sampled human labels over time. Here is the limitation to watch for: A stable average occupancy count could still hide a shifted camera. The answer follows from that evidence: Drift monitoring tracks changes in viewpoints, lighting, data, and error rates. Compare landmark positions before and after the camera tilt. A changed space map may look like a detector failure when the underlying boxes are sound. Keep timestamped frames so affected counts can be identified. Compare with a reference view taken before maintenance to locate the shift.
Next, we'll plan how to update the vision system without letting a well-intended fix quietly introduce new errors.