This page is located in archive. Go to the latest version of this course pages.

Requirements for the final report

This is currently obsolete and not valid….

The final report need not be written as full technical report, we need only documentation of your results.

Required Content

  • Header: title (including the code of the subject – ..TDV), author, date.
  • Task overview. Briefly describe the task you are solving (the task is to create a 3D model, not to practice something), describe the origin of the data. Give a basic overview of the task (can be structured), but do not describe the standard methods. Describe a method only if some invention is made (methods proposed or modified by authors).
  • Data and results. Document all your data, parameters of algorithms and intermediate and final results.
  • Conclusion, summary, discussion. Rating of the results. Some more ideas or proposals, if there are any.

Keep the text brief. It is OK when only standard methods are used and there is nothing more to be written.

Technical content

All facts, intermediate results and the final result must be documented. The (intermediate) results should also demonstrate that the solution of the problem is correct. The list of facts that should be present in the report includes:

  • Used data parameters
    • The images: what is the object, sizes of images, approximate distance between capturing locations, specification of the camera. Number the images allowing to refer them.
    • Thresholds for RANSAC.
  • Quantitative results
    • Radial distortion calibration accuracy.
    • Camera calibration matrix K (undistorted images).
    • Counts of pair-wise correspondences (tentative and inliers), reprojection errors (all used pairs).
    • Counts of RANSAC iterations.
    • Sequence of cameras during step-wise gluing.
    • Counts of image-to-scene correspondences (tentative, inliers), reprojection errors (all images)
    • Number of points in sparse 3D point cloud.
    • Which pairs were used in stereo-matching.
    • Number of points in dense 3D point cloud.
  • Qualitative results
    • All input images, numbered for later referencing.
    • Scheme of capturing, including (approximate) distances between cameras.
    • Images used for internal calibration.
    • Image of pair-wise correspondences in a pair that initialises step-wise gluing. Distinguish inliers and outliers.
    • Illustration of epipolar geometry found on the pair (few corresponding points and their epipolar lines in both images).
    • 3D plot of sparse point cloud together with the cameras.
    • Example of epipolarly rectified images (choose some nice pair).
    • Disparity maps (more, better all of them, put more alongside on a single page.
    • Picture of a reconstructed surface.

Formal requirements

The report can be written in English, Czech or Slovak. Please use page numbering, A4 page format, font size 10-12 and spellchecker. Do not create separate title page and do not include any empty pages. Do not create table of contents. Do not enforce page-break before the chapters and sections. The text need not be continuous text only, parts can be also expressed in structured way (e.g. lists).

Every figure and table must contain a caption with brief description – what is there, what should be seen, legend (explaining colours, line styles, and so).

Do not describe the implementation, write about principles. Try to express them in a mathematical way. Implementation details should be mentioned only in reasonable cases, e.g., we do some minimisation (x=argmin(f)), and we implement it using Matlab function fminsearch.

The (brief) text should not be written in the colloquial strain. The result use language of a technical report, not of a tale. Except in final discussion, avoid speculations, but write and document specifically what you have done (and what haven't). Alternatives can be mentioned only if there is an explanation which one is chosen and why. Either active or passive voice can be used, but it should be consistent over whole text.

Note for reports written in Czech: avoid slang, write in Czech, not in 'Czenglish'. Special hard-to-translate English terms can be used ('inliery', 'outliery'), but use Czech words if they exist (e.g. match → pár, korespondence).


The work must be your own. We have zero tolerance of plagiarism.

Certainly, the work uses statements and facts that are not results of author's work (e.g. use of some standard method or available implementation). Usually, it is inadmissible to include a verbatim copy of somebody else's work (including images, etc.) without permission.

courses/tdv/labs/report_requirements.txt · Last modified: 2018/10/01 10:58 (external edit)