Watch for Small-Baseline Failure
Both cameras are almost in the same position.
Watch for Small-Baseline Failure explained visually
Both cameras are almost in the same position. English captions are optional and start off. The complete transcript follows the written lesson.
What you will learn
Both cameras are almost in the same position.
A practical question
Two cameras are mounted almost side by side, looking at a distant parked car. The corresponding viewing rays are nearly parallel. A tiny error in placing one image point can now shift the estimated distance dramatically.
The useful idea
The separation between the cameras is the baseline. A larger useful baseline gives more viewpoint difference for distant depth, though it can make matching harder when objects appear very different or one view hides a surface. Camera calibration and pixel precision also affect the result.
Work through the example
Compare triangulation on a nearby cone and a far wall with the same camera pair. Perturb one marked pixel slightly and watch how the depth estimate changes. That sensitivity is an honest uncertainty, not something to hide by rounding the number.
Limits and next step
Choose a setup suited to the distances you need and report where it becomes unreliable. Next, we'll express the stereo depth relationship through disparity in a rectified pair.
Worked example
Compare depth estimates from a short and a larger baseline for the same distant point.
Quick check: test your understanding
Try it yourself: In the situation described above, explain the method and one reason its output might be misleading.
Check the reasoning
A tiny baseline gives weak depth parallax and amplifies triangulation uncertainty. A larger baseline can improve geometry but make matching and occlusion harder.
What is the main idea behind watch for small-baseline failure?
A tiny baseline gives weak depth parallax and amplifies triangulation uncertainty.
How is it used in the example?
Compare depth estimates from a short and a larger baseline for the same distant point.
What limitation should be checked?
A larger baseline can improve geometry but make matching and occlusion harder.
Further reading
Programming Computer Vision with Python — Jan Erik Solem.
Original ToolsInVision teaching examples. Editorial approach · Suggest a correction
Accessible referenceLesson 36 transcript150 words · Video transcript · Matches the published narration and captionsOpen transcript
Two cameras are mounted almost side by side, looking at a distant parked car. The corresponding viewing rays are nearly parallel. A tiny error in placing one image point can now shift the estimated distance dramatically.
The separation between the cameras is the baseline. A larger useful baseline gives more viewpoint difference for distant depth, though it can make matching harder when objects appear very different or one view hides a surface. Camera calibration and pixel precision also affect the result.
Compare triangulation on a nearby cone and a far wall with the same camera pair. Perturb one marked pixel slightly and watch how the depth estimate changes. That sensitivity is an honest uncertainty, not something to hide by rounding the number.
Choose a setup suited to the distances you need and report where it becomes unreliable. Next, we'll express the stereo depth relationship through disparity in a rectified pair.