Improve Virtual Stereo Calibration #1

Myzhar · 2025-11-11T17:55:32Z

Improve the experience with Virtual Stereo Calibration
Add ZED SDK compatible calibration files

Copilot

Pull Request Overview

This PR significantly improves the virtual stereo calibration experience by introducing a new calibration quality checker system and adding support for ZED SDK-compatible calibration file formats. The changes enhance the user experience with better visual feedback and more robust sample collection criteria.

Key Changes

Implemented a new CalibrationChecker class that evaluates calibration sample quality based on position, size, and skew metrics
Added ZED SDK format calibration file export in addition to OpenCV format
Enhanced UI with real-time coverage indicators and better status messages for sample acquisition

Reviewed Changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 14 comments.

Show a summary per file

File	Description
stereo_calibration/src/opencv_calibration.cpp	Refactored calibration function with improved error handling and added ZED SDK format file generation
stereo_calibration/include/opencv_calibration.hpp	Updated function signatures to support new parameters (is_dual_mono, is_4k, verbose) and changed matrix type from CV_64FC1 to CV_32FC1
stereo_calibration/src/main.cpp	Overhauled UI/UX with calibration checker integration, improved argument parsing, and enhanced visual feedback
stereo_calibration/src/calibration_checker.cpp	New file implementing sample quality evaluation based on checkerboard position, size, and skew
stereo_calibration/include/calibration_checker.hpp	New header defining the CalibrationChecker class and related data structures
stereo_calibration/.vscode/settings.json	VS Code configuration file disabling error squiggles

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

stereo_calibration/include/opencv_calibration.hpp

stereo_calibration/src/opencv_calibration.cpp

Copilot · 2025-11-20T16:20:48Z

stereo_calibration/src/opencv_calibration.cpp

+  std::cout
+      << "The resolution for the calibration is not valid\n\nUse 4k "
+         "(3840x2160) for ZED X One 4K and FHD1200 (1920x1200) for ZED X One GS"
+      << std::endl;
+
+  return std::string();


The control flow will never reach these lines because the function returns at line 409 for the 4K camera case (is_4k == true) and at line 301 for the non-4K case (is_4k == false). This unreachable code should be removed or the function logic should be restructured.

Additionally, the error message is misleading since it will never be displayed.

Suggested change

std::cout

<< "The resolution for the calibration is not valid\n\nUse 4k "

"(3840x2160) for ZED X One 4K and FHD1200 (1920x1200) for ZED X One GS"

<< std::endl;

return std::string();

stereo_calibration/src/calibration_checker.cpp

stereo_calibration/src/opencv_calibration.cpp

stereo_calibration/src/main.cpp

stereo_calibration/src/calibration_checker.cpp

stereo_calibration/include/calibration_checker.hpp

Co-authored-by: Copilot <[email protected]>

Copilot

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 20 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2025-11-28T09:16:41Z

stereo_calibration/src/main.cpp

+const float min_y_coverage =
+    0.7f;  // Checkerboard Y position should cover 70% of the image height
+const float min_area_range = 0.45f;  // Checkerboard area range size should be at
+                                    // least 0.4 [min_area-max_area]


Typo in comment at line 18: "at least 0.4 [min_area-max_area]" but the actual value is 0.45f. The comment should match the constant value or vice versa. Update to: // least 0.45 [min_area-max_area]

Suggested change

// least 0.4 [min_area-max_area]

// least 0.45 [min_area-max_area]

Copilot · 2025-11-28T09:16:42Z

stereo_calibration/src/calibration_checker.cpp

+  // min_skew = 0.0f;
+  // min_size = 0.0f;


[nitpick] The commented-out code at lines 299-300 should be removed. If there's a reason to keep the option of not rewarding small size or skew in the future, this should be documented or made configurable rather than left as commented code.

Suggested change

// min_skew = 0.0f;

// min_size = 0.0f;

Copilot · 2025-11-28T09:16:42Z

stereo_calibration/src/main.cpp

+  if (top_left_count >= min_samples / 4) {
+    cv::rectangle(
+        coverage_indicator, cv::Point(0, 0),
+        cv::Point(coverage_indicator.cols / 2, coverage_indicator.rows / 2),
+        cv::Scalar(255), -1);
+  }
+  if (top_right_count >= min_samples / 4) {
+    cv::rectangle(
+        coverage_indicator, cv::Point(coverage_indicator.cols / 2, 0),
+        cv::Point(coverage_indicator.cols, coverage_indicator.rows / 2),
+        cv::Scalar(255), -1);
+  }
+  if (bottom_left_count >= min_samples / 4) {
+    cv::rectangle(
+        coverage_indicator, cv::Point(0, coverage_indicator.rows / 2),
+        cv::Point(coverage_indicator.cols / 2, coverage_indicator.rows),
+        cv::Scalar(255), -1);
+  }
+  if (bottom_right_count >= min_samples / 4) {


[nitpick] Potential integer division issue: min_samples / 4 at lines 575, 581, 587, 593 performs integer division. If min_samples is 25 (as defined), then min_samples / 4 = 6 (with remainder discarded). This means the four quadrants require 6 samples each for a total of 24, not 25. Consider documenting this behavior or using a more explicit calculation like (min_samples + 3) / 4 to handle rounding more explicitly.

Suggested change

if (top_left_count >= min_samples / 4) {

cv::rectangle(

coverage_indicator, cv::Point(0, 0),

cv::Point(coverage_indicator.cols / 2, coverage_indicator.rows / 2),

cv::Scalar(255), -1);

}

if (top_right_count >= min_samples / 4) {

cv::rectangle(

coverage_indicator, cv::Point(coverage_indicator.cols / 2, 0),

cv::Point(coverage_indicator.cols, coverage_indicator.rows / 2),

cv::Scalar(255), -1);

}

if (bottom_left_count >= min_samples / 4) {

cv::rectangle(

coverage_indicator, cv::Point(0, coverage_indicator.rows / 2),

cv::Point(coverage_indicator.cols / 2, coverage_indicator.rows),

cv::Scalar(255), -1);

}

if (bottom_right_count >= min_samples / 4) {

if (top_left_count >= (min_samples + 3) / 4) {

cv::rectangle(

coverage_indicator, cv::Point(0, 0),

cv::Point(coverage_indicator.cols / 2, coverage_indicator.rows / 2),

cv::Scalar(255), -1);

}

if (top_right_count >= (min_samples + 3) / 4) {

cv::rectangle(

coverage_indicator, cv::Point(coverage_indicator.cols / 2, 0),

cv::Point(coverage_indicator.cols, coverage_indicator.rows / 2),

cv::Scalar(255), -1);

}

if (bottom_left_count >= (min_samples + 3) / 4) {

cv::rectangle(

coverage_indicator, cv::Point(0, coverage_indicator.rows / 2),

cv::Point(coverage_indicator.cols / 2, coverage_indicator.rows),

cv::Scalar(255), -1);

}

if (bottom_right_count >= (min_samples + 3) / 4) {

Copilot · 2025-11-28T09:16:42Z

stereo_calibration/src/main.cpp

+
+constexpr int target_w = 9;          // number of horizontal inner edges
+constexpr int target_h = 6;          // number of vertical inner edges
+constexpr float square_size = 25.4;  // mm


[nitpick] The default checkerboard square size is 25.4mm (1 inch), but the comment says "mm". This is correct but could be more explicit that 25.4mm equals 1 inch for the standard OpenCV checkerboard pattern. Consider adding: // mm (1 inch for the standard OpenCV pattern)

Suggested change

constexpr float square_size = 25.4; // mm

constexpr float square_size = 25.4; // mm (1 inch for the standard OpenCV pattern)

Copilot · 2025-11-28T09:16:43Z

stereo_calibration/src/opencv_calibration.cpp

+      std::cout << " !!! Failed to save OpenCV calibration file " << opencv_file
+                << " !!!" << std::endl;


[nitpick] Inconsistent spacing in the error message. Line 161 has " !!! Failed to save OpenCV calibration file " with spaces around the file name, but line 172 has the same message for ZED SDK format. Both should have consistent formatting.

Copilot · 2025-11-28T09:16:46Z

stereo_calibration/src/opencv_calibration.cpp

+    outfile << "cx = " << left.K.at<float>(0, 2) - (3840 - 1920) / 2
+            << "\n";
+    outfile << "cy = " << left.K.at<float>(1, 2) - (2160 - 1200) / 2
+            << "\n\n";
+
+    outfile << "[RIGHT_CAM_FHD1200]\n";
+    outfile << "fx = " << right.K.at<float>(0, 0) << "\n";
+    outfile << "fy = " << right.K.at<float>(1, 1) << "\n";
+    outfile << "cx = " << right.K.at<float>(0, 2) - (3840 - 1920) / 2
+            << "\n";
+    outfile << "cy = " << right.K.at<float>(1, 2) - (2160 - 1200) / 2


[nitpick] The division operation (3840 - 3200) / 2 at lines 330, 338, 358, 366 uses integer division which results in 320. Similarly, (2160 - 1800) / 2 results in 180, and (3840 - 1920) / 2 results in 960, and (2160 - 1200) / 2 results in 480. While these are correct, the float subtraction context might cause confusion. Consider adding explicit comments or using floating-point division if needed for clarity.

Suggested change

outfile << "cx = " << left.K.at<float>(0, 2) - (3840 - 1920) / 2

<< "\n";

outfile << "cy = " << left.K.at<float>(1, 2) - (2160 - 1200) / 2

<< "\n\n";

outfile << "[RIGHT_CAM_FHD1200]\n";

outfile << "fx = " << right.K.at<float>(0, 0) << "\n";

outfile << "fy = " << right.K.at<float>(1, 1) << "\n";

outfile << "cx = " << right.K.at<float>(0, 2) - (3840 - 1920) / 2

<< "\n";

outfile << "cy = " << right.K.at<float>(1, 2) - (2160 - 1200) / 2

// Subtract half the difference to center for FHD1200 crop

outfile << "cx = " << left.K.at<float>(0, 2) - (3840.0f - 1920.0f) / 2.0f

<< "\n";

outfile << "cy = " << left.K.at<float>(1, 2) - (2160.0f - 1200.0f) / 2.0f

<< "\n\n";

outfile << "[RIGHT_CAM_FHD1200]\n";

outfile << "fx = " << right.K.at<float>(0, 0) << "\n";

outfile << "fy = " << right.K.at<float>(1, 1) << "\n";

// Subtract half the difference to center for FHD1200 crop

outfile << "cx = " << right.K.at<float>(0, 2) - (3840.0f - 1920.0f) / 2.0f

<< "\n";

outfile << "cy = " << right.K.at<float>(1, 2) - (2160.0f - 1200.0f) / 2.0f

Copilot · 2025-11-28T09:16:46Z

stereo_calibration/src/calibration_checker.cpp

+        std::abs(p1.pos.x - p2.pos.x) / std::max(std::max(p1.pos.x, p2.pos.x), epsilon);
+    float pos_y_diff =
+        std::abs(p1.pos.y - p2.pos.y) / std::max(std::max(p1.pos.y, p2.pos.y), epsilon);
+    float size_diff = std::abs(p1.size - p2.size) / std::max(std::max(p1.size, p2.size), epsilon);
+    float skew_diff = std::abs(p1.skew - p2.skew) / std::max(std::max(p1.skew, p2.skew), epsilon);


The division by std::max(std::max(p1.pos.x, p2.pos.x), epsilon) at lines 227 and 229 can produce incorrect results when both values are close to zero. For example, if p1.pos.x = 0.001 and p2.pos.x = 0.002, the difference is 0.001 and the max is 0.002, giving a diff of 0.001/0.002 = 0.5 = 50%, which seems incorrect for such small absolute values. Consider using the absolute difference relative to a meaningful scale (like 1.0) or the average of the two values instead of the maximum.

Suggested change

std::abs(p1.pos.x - p2.pos.x) / std::max(std::max(p1.pos.x, p2.pos.x), epsilon);

float pos_y_diff =

std::abs(p1.pos.y - p2.pos.y) / std::max(std::max(p1.pos.y, p2.pos.y), epsilon);

float size_diff = std::abs(p1.size - p2.size) / std::max(std::max(p1.size, p2.size), epsilon);

float skew_diff = std::abs(p1.skew - p2.skew) / std::max(std::max(p1.skew, p2.skew), epsilon);

std::abs(p1.pos.x - p2.pos.x) / std::max((std::abs(p1.pos.x) + std::abs(p2.pos.x)) / 2.0f, epsilon);

float pos_y_diff =

std::abs(p1.pos.y - p2.pos.y) / std::max((std::abs(p1.pos.y) + std::abs(p2.pos.y)) / 2.0f, epsilon);

float size_diff = std::abs(p1.size - p2.size) / std::max((std::abs(p1.size) + std::abs(p2.size)) / 2.0f, epsilon);

float skew_diff = std::abs(p1.skew - p2.skew) / std::max((std::abs(p1.skew) + std::abs(p2.skew)) / 2.0f, epsilon);

Copilot · 2025-11-28T09:16:47Z

stereo_calibration/src/main.cpp

+        std::cerr << "Error: Left and Right camera IDs or Left and Right "
+                     "camera Serial Numbers must be both provided."
+                  << std::endl;
+        std::cerr << " * use the command " << args.app_name


The error message uses single quotes around the command but the actual command string has a typo - there's an extra single quote: " * use the command " << args.app_name << " -h' for details.". Should be: " * use the command '" << args.app_name << " -h' for details."

Suggested change

std::cerr << " * use the command " << args.app_name

std::cerr << " * use the command '" << args.app_name

Copilot · 2025-11-28T09:16:47Z

stereo_calibration/src/opencv_calibration.cpp

+    outfile << "k3 = " << calib.D.at<float>(2) << "\n";
+    outfile << "k4 = " << calib.D.at<float>(3) << "\n";
+  }
+  outfile<<"\n";


[nitpick] Missing newline character in output. The message on line 222 (outfile<<"\n";) should consistently use either "\n" or std::endl. The rest of the function uses "\n" inside string literals, so this standalone statement is inconsistent. Consider using outfile << "\n"; if needed, or remove it if the newline is already included in the previous line.

Suggested change

outfile<<"\n";

Copilot · 2025-11-28T09:16:47Z

stereo_calibration/src/opencv_calibration.cpp

+    std::cerr
+        << " !!! Cannot save the calibration file: 'Unable to open output file'"
+        << std::endl;
+    return std::string();


[nitpick] The duplicate message "Parameter file written successfully" appears on both line 303 and line 395. This is correct behavior (one for each camera type), but the success message is printed before returning, while the error path (lines 231-234) returns an empty string. However, the error message on line 232 says "Unable to open output file" but this could also happen at line 308 or 313 where the resolution validation fails - those don't print this message. Consider consolidating error handling or making error messages more specific.

…encv-calibration into improve_virtual_stereo

P-yver · 2025-12-02T19:26:01Z

stereo_calibration/src/opencv_calibration.cpp

+      return EXIT_FAILURE;
+    }
+
+    if (calib_data.T.at<float>(0) > 0) {


duplicated check

Add parameters for Virtual Stereo init

f69e462

Myzhar self-assigned this Nov 11, 2025

Myzhar added 20 commits November 11, 2025 19:28

Improve result report

51404dc

Improve calibration feedback

8847f0e

Fix image reading error

09e149d

Debug info

529a117

Improve target detection

91eaad3

Add coverage Checker code

cb92b29

Test calibration checker

7a40e46

Test sample collection completeness

14b6b4e

Calibration validated

cf522f5

Change params diff check

5342e6a

Better user feedback

5b9ea50

Fix sample diff checker

d462be5

TODO Check sample differences

9131b8c

Add 4-zone coverage indicator

110f7cd

Check ZED X One model

db871e4

Add function to save SDK conf file

309b34d

Save SDK calib file

8dd778a

Fix Rotation vector for SDK calib file

4f2108b

Minor Fix

aa2f42d

Improve skew calc and fix SDK calib saving

f5430ee

Myzhar requested a review from Copilot November 20, 2025 16:14

Copilot started reviewing on behalf of Myzhar November 20, 2025 16:14 View session

Copilot finished reviewing on behalf of Myzhar November 20, 2025 16:16

Copilot AI reviewed Nov 20, 2025

View reviewed changes

Myzhar and others added 4 commits November 20, 2025 17:22

Update stereo_calibration/src/opencv_calibration.cpp

39d913d

Co-authored-by: Copilot <[email protected]>

Update stereo_calibration/src/calibration_checker.cpp

1c82ed5

Co-authored-by: Copilot <[email protected]>

Fixes by Copilot review

e521ca3

Data type fix and param tuning

467e127

fix fisheye disto export

6b0a943

Myzhar requested a review from Copilot November 28, 2025 09:10

Copilot started reviewing on behalf of Myzhar November 28, 2025 09:10 View session

P-yver marked this pull request as ready for review November 28, 2025 09:12

Copilot finished reviewing on behalf of Myzhar November 28, 2025 09:13

Copilot AI reviewed Nov 28, 2025

View reviewed changes

Myzhar added 5 commits December 2, 2025 12:59

Merge branch 'improve_virtual_stereo' of github.com:stereolabs/zed-op…

8abe278

…encv-calibration into improve_virtual_stereo

Add baseline value check

c74a81c

Add checkerboard params

9e88f8d

Improve user log feedback

ce2c0f6

Improve user feedback

9ec055f

P-yver reviewed Dec 2, 2025

View reviewed changes

stereo_calibration/src/opencv_calibration.cpp Outdated

return EXIT_FAILURE;

}

if (calib_data.T.at<float>(0) > 0) {

Copy link

Member

P-yver Dec 2, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

duplicated check

Myzhar added 2 commits December 2, 2025 20:46

Improvements

7781ce4

Added minimum checkerboard area param

ea5c491

	// least 0.4 [min_area-max_area]
	// least 0.45 [min_area-max_area]

	constexpr float square_size = 25.4; // mm
	constexpr float square_size = 25.4; // mm (1 inch for the standard OpenCV pattern)

		std::cout << " !!! Failed to save OpenCV calibration file " << opencv_file
		<< " !!!" << std::endl;

	std::cerr << " * use the command " << args.app_name
	std::cerr << " * use the command '" << args.app_name

Improve Virtual Stereo Calibration #1

Are you sure you want to change the base?

Improve Virtual Stereo Calibration #1

Uh oh!

Conversation

Myzhar commented Nov 11, 2025

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Key Changes

Reviewed Changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI Nov 20, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Nov 28, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 28, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 28, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 28, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 28, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 28, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 28, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 28, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 28, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 28, 2025

Choose a reason for hiding this comment

Uh oh!

P-yver Dec 2, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

3 participants