Different ID?

[samxuxiang]

Hi, I wonder if the file id (e.g. 20440_27177360_0004) is the same for reconstruction and segmentation data? There are around 4500 cad models with the same id, but not all of them point to the same object. Is there any way to find the correct overlapping objects in the two datsets? Thank you.

[karldd]

For the reconstruction dataset the naming convention is as follows: XXXXX_YYYYYYYY_ZZZZ[_1234].ext.

• XXXXX represents the project
• YYYYYYYY the file
• ZZZZ the component
• _1234 the extrude index

If _1234 is absent the file represents the final design rather than a step in the sequence.

XXXXX_YYYYYYYY is common between the two datasets, and I believe ZZZZ is also the same, but with different zero padding.

However, the designs themselves differ because they suppress different modeling operations. Suppressing fillets, for example looks like this:

The design on the left is the original design, and the design on the right is the design without fillets and only sketch and extrude operations.

The segmentation dataset has metadata from: extrude, chamfer, fillet, and revolve. The reconstruction dataset has the full construction sequence for sketch and extrude.

So if you were too try and find the common CAD models, you could start by matching the file names, then do geometric checks to see if the design actually matches or if the feature suppression has changed the design. I'm not sure of the best way to do this however.

Is there a specific goal you have in mind with finding the common CAD models?

[samxuxiang]

Thank you for the reply. The issue I run into is a bit different. For example, the cad model 20440_27177360_0004 in reconstruction looks like this:

But the corresponding model 20440_27177360_4 in segmentation looks like this:

[karldd]

XXXXX_YYYYYYYY is common between the two datasets, and I believe ZZZZ is also the same, but with different zero padding.

I see. Yes this seems to confirm that the ZZZZ is in fact different between the two datasets.

Another approach could be to try and map the following json:

Reconstruction dataset for 20440_27177360_0004

    "metadata": {
        "parent_project": "20440_27177360",
        "component_name": "tampo v3",
        "component_index": 4
    },

Segmentation dataset under timeline_info for 20440_27177360_4

    "metadata": {
        "component_name": "pino v2"
    },

So you should be able to use the component_name to match them as this should be unique under XXXXX_YYYYYYYY. Just keep in mind that:

1. There is no guarantee the same files with exist in both datasets due to different processing pipelines
2. The 3D shapes may be different if the files do exists due to modeling operation suppression

[samxuxiang]

Thank you!