alicevision / meshroom Goto Github PK

In the right click menu, we can "open parent directory" of an image which is useful.
But when the job is started everything is locked. And we cannot retrieve the folder of the input images from meshroom.

Visualize absolute paths of source images.
Maybe also a right click to open the folder directly.

Could be very useful to switch between full view and thumbnail view mode by double click.

By default don't set the image pair and rely on the openMVG automatic selection.

As we will be able to add images, we need to support the notion of versions in the application.

load the Alembic file, when available, when selecting the AlembicExport node.
clear the 3d view otherwise.

job_start.py --engine=local JOB.JSON
job_start.py --engine=dummy JOB.JSON
job_start.py --engine=tractor JOB.JSON (the default value)

We need to be able to launch a reconstruction with SIFT only, CCTag only or SIFT+CCTag.

The file argument is:

"feature_detection": {
    "method": "SIFT_CCTAG3"
}

Possible values are:

• SIFT (default)
• CCTAG3: CCTAG markers with 3 crowns
• SIFT_CCTAG3

• the project name isn't immediately updated on the homepage, after a rename
• sometimes, we can't delete an image from the list (keyboard focus issue)
• show the image count per job
• change the status color & title size
• add an 'import image...' menu
• add some parameter tooltips
• submit on farm, by default
• the 'plus' button should add a new empty job
• remove the job name from the main menuBar
• allow a drag n'drop of the initial pair (from an external image browser)
• avoid the qml warning when an image is unavailable
• avoid the qml warnings on window destruction
• add a visible scrollbar to the image grid/list view

Currently the name is retrieved from the selected folder, but in the pipeline it will always be "photogrammetry".
So it could be great if the user could customize it.

The underlying principle of job submission in meshroom can be represented as a graph of different stages to perform, like feature extraction then feature matching then sfm, and so on. For now this graph is expressed in the scriptMVG part of the project. Investigate what it would take to have this computation graph directly in meshroom to eventually manipulate it directly in the UI.

• Design doc
• Small proof of concept prototype

• Visualize the reliability of the cameras with a color range (green is strong & red is weak). We can see this color on the 3D gizmo as well as in 2D as a border of the thumbnail.
• Add a slider to filter the cameras based on this reliability.

For this first step, the camera reliability will be defined by the number of reconstructed points in each camera.

Need a button to open the folder containing all the output files.

replace the custom 3d view with a Qt3d-qml viewport

When we use the thumbnail mode, we can select and remove images.
It could be very useful to be able to remove the current image when we are in full image view mode.

• local computation
• no REZ

Switch between images from the keyboard (left/right keys) in full image view mode.

We probably want to add all the images inside.

refresh all node statuses:

• when a new scene is loaded
• on demand (user action)

We need a 3d viewer in mockupMVG. In the first version this viewer should display colored point clouds, camera positions and orientation. We should be able to move around in the 3d space using the maya mouse interaction.

Only keep JPG/jpg/JPEG/jpeg files.

Run & visualize tracking of video files without RT constraint.
Should be able to compute on renderfarm.

use Qt5.7, with Qt Quick Controls 2.0 and the new stylesheet (QQuickStyle) system

mouse controls used in the GL view should work exactly as in Autodesk Maya.
these controls are sometimes inverted :

• horizontal pan direction, after rotating the camera
• zoom in/out direction, using the mouse wheel

• Visualize the reliability of the 3D points with a color range (green is strong & red is weak).
• Add a slider to filter the point cloud based on this reliability.

For this first step, the 3D points reliability will be defined by the visibility size.

mimic maya camera controls with qt3d

The text zone always keep the focus.

We may need to add a global background that accept focus, like that:
https://github.com/buttleofx/QuickMamba/blob/master/qml/QuickMamba/TopFocusHandler.qml

• option to disable all cameras
• option to change the scale of the locators

replace the old importer class with a new Qt3d module

Maybe Ctrl+A to select all images so we could remove them globally.

Currently we can only see thumbnails, so it's difficult to remove images, select the initial pair.

• Simple documentation on how to use the soft.
  • French
  • English
  • Use nice icons
• FAQ to explain the common errors

http://fr.eyefi.com

Add metadata to an image from meshroom.

• ensure that everything is fine on linux
• temporarily remove the theia dependency

As in meshlab on double click or as in Maya with F.

Require picking.
Require selection if we do the same than in Maya.

A Mikros document already exists, but it needs to be improved.

Choose a custom name for reconstructions.
Could be also useful to have a way to put a long description.

See poparteu/scriptMVG#5

For tracking we don't need meshing or undistort images.

File format supported by scriptMVG-dev_mikros is:

    "steps": {
        "meshing": {
            "enabled": 0,
            "scale": 2
        },
        "undistort": {
            "enabled": 0
        },

Duplicate a job to launch the same job with different parameters.
At least copy the images.

"Job > Import Images" open a popup with a filter to list only images (instead of all files). But it doesn't list JPG files.