{ "id": 508768, "title": "Statement on Issues for Phase 2 Data", "url": "/competitions/cvpr-metafood-3d-food-reconstruction-challenge/discussion/508768", "author_name": "", "post_date": "2024-05-31T03:06:00.048454600Z", "votes": 1, "comment_count": 7, "views": 0, "content": "

Dear Phase 2 Participants,

\n

Both Team VoIETA and VCE have raised issues with our dataset regarding incorrect meshes, scaling issues, and code related to point cloud transformations.

\n
    \n

  1. Incorrect Meshes for Scene 12 (steak) and Scene 15 (Chicken nugget): We have traced back to our large dataset but were unable to find the original mesh scans for these two scenes. Therefore, we regret to inform you that we must remove these two scenes from all score calculations. We apologize for the inconvenience this may cause and appreciate your understanding.

    2. \n

  2. Scaling Issues: The first version of Phase 2 meshes we uploaded were not correctly scaled. We have re-uploaded the ground truth meshes in the ground_truth_scaled folder. Please use these for your final transformation calculations.

    3. \n

  3. Point Cloud Transform Discrepancy: Team VoIETA found a discrepancy in the point clouds after applying the transform in our script. However, our tests did not reveal this issue. Team VoIETA has provided a code snippet for their transformation function using Open3D. We will adapt their code and evaluate the results based on both our original transformation function and the Open3D one provided by Team VoIETA. We will use the minimum distance from the two as the final result. See the Phase 2 Data Are Now Available post for more details.

    4. \n
\n

Thank you for your understanding and participation in the challenge. We apologize for any inconvenience caused.

\n

Sincerely,
\nMetaFood Workshop Organizing Committee

", "messages": [ { "id": "2846225", "postDate": "05/31/2024 03:06:00", "content": "

Dear Phase 2 Participants,

\n

Both Team VoIETA and VCE have raised issues with our dataset regarding incorrect meshes, scaling issues, and code related to point cloud transformations.

\n
    \n

  1. Incorrect Meshes for Scene 12 (steak) and Scene 15 (Chicken nugget): We have traced back to our large dataset but were unable to find the original mesh scans for these two scenes. Therefore, we regret to inform you that we must remove these two scenes from all score calculations. We apologize for the inconvenience this may cause and appreciate your understanding.

    2. \n

  2. Scaling Issues: The first version of Phase 2 meshes we uploaded were not correctly scaled. We have re-uploaded the ground truth meshes in the ground_truth_scaled folder. Please use these for your final transformation calculations.

    3. \n

  3. Point Cloud Transform Discrepancy: Team VoIETA found a discrepancy in the point clouds after applying the transform in our script. However, our tests did not reveal this issue. Team VoIETA has provided a code snippet for their transformation function using Open3D. We will adapt their code and evaluate the results based on both our original transformation function and the Open3D one provided by Team VoIETA. We will use the minimum distance from the two as the final result. See the Phase 2 Data Are Now Available post for more details.

    4. \n
\n

Thank you for your understanding and participation in the challenge. We apologize for any inconvenience caused.

\n

Sincerely,
\nMetaFood Workshop Organizing Committee

", "rawMarkdown": "Dear Phase 2 Participants,\n\nBoth Team VoIETA and VCE have raised issues with our dataset regarding incorrect meshes, scaling issues, and code related to point cloud transformations.\n\n1. **Incorrect Meshes for Scene 12 (steak) and Scene 15 (Chicken nugget):** We have traced back to our large dataset but were unable to find the original mesh scans for these two scenes. Therefore, we regret to inform you that we must remove these two scenes from all score calculations. We apologize for the inconvenience this may cause and appreciate your understanding.\n\n2. **Scaling Issues:** The first version of Phase 2 meshes we uploaded were not correctly scaled. We have re-uploaded the ground truth meshes in the `ground_truth_scaled` folder. Please use these for your final transformation calculations.\n\n3. **Point Cloud Transform Discrepancy:** Team VoIETA found a discrepancy in the point clouds after applying the transform in our script. However, our tests did not reveal this issue. Team VoIETA has provided a code snippet for their transformation function using Open3D. We will adapt their code and evaluate the results based on both our original transformation function and the Open3D one provided by Team VoIETA. We will use the minimum distance from the two as the final result. See the Phase 2 Data Are Now Available post for more details. \n\nThank you for your understanding and participation in the challenge. We apologize for any inconvenience caused.\n\nSincerely,\nMetaFood Workshop Organizing Committee", "votes": null }, { "id": "2847926", "postDate": "05/31/2024 18:08:29", "content": "

Dear Organizers,

\n

We are glad that our suggestion has been accepted. Yet, we found some issues with the new meshes:

\n\n
Mesh is not , no information on volume, barycenter and inertia tensor.\n
\n\n
RuntimeError: [Open3D Error] (std::shared_ptr<open3d::geometry::PointCloud> open3d::geometry::TriangleMesh::SamplePointsPoissonDisk(size_t, double, std::shared_ptr<open3d::geometry::PointCloud>, bool)) /root/Open3D/cpp/open3d/geometry/TriangleMesh.cpp:527: Input mesh has no triangles.\n
\n

", "rawMarkdown": "Dear Organizers, \n\nWe are glad that our suggestion has been accepted. Yet, we found some issues with the new meshes:\n\n* Some meshes are not watertight (e.g., mesh number 1), so we cannot see the actual volumes to validate phase 1. Example error from MeshLab:\n```bash\nMesh is not 'watertight', no information on volume, barycenter and inertia tensor.\n```\n* Mesh number 6 has no triangles, which is not parsable like other meshes. Example error:\n```bash\nRuntimeError: [Open3D Error] (std::shared_ptr open3d::geometry::TriangleMesh::SamplePointsPoissonDisk(size_t, double, std::shared_ptr, bool)) /root/Open3D/cpp/open3d/geometry/TriangleMesh.cpp:527: Input mesh has no triangles.\n```\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8304568%2Ffc360ec7775552006f7542042b452f45%2FScreenshot%20from%202024-05-31%2020-07-28.png?generation=1717178865781436&alt=media)", "votes": null }, { "id": "2848029", "postDate": "05/31/2024 18:51:50", "content": "

Thanks for pointing these out!

\n
    \n
  1. While some of the meshes may not be watertight, we are still able to obtain the volume of these meshes and that is what is used in our calculation. We use trimesh python library for this volume calculation. Additionally, you can use the 3D print toolbox add-on in Blender and obtain the volume of the mesh too.
    2. \n
  2. Further, we are able to open the 6.obj file on all of our software and libraries including Windows 3D Viewer, Blender, MeshLab, and Trimesh. We haven't checked with Open3D but the format of the vertices, faces, normals and other information in the obj file is consistent with the rest so you shouldn't be facing any issues there. A quick suggestion could be redownloading the files and trying to open them without any modifications for sanity check. If it still fails to open, you can try other software or libraries to check if the issue persists.
    3. \n
\n

Please let us know if there are further questions, thank you!

", "rawMarkdown": "Thanks for pointing these out! \n1. While some of the meshes may not be watertight, we are still able to obtain the volume of these meshes and that is what is used in our calculation. We use trimesh python library for this volume calculation. Additionally, you can use the 3D print toolbox add-on in Blender and obtain the volume of the mesh too.\n2. Further, we are able to open the 6.obj file on all of our software and libraries including Windows 3D Viewer, Blender, MeshLab, and Trimesh. We haven't checked with Open3D but the format of the vertices, faces, normals and other information in the obj file is consistent with the rest so you shouldn't be facing any issues there. A quick suggestion could be redownloading the files and trying to open them without any modifications for sanity check. If it still fails to open, you can try other software or libraries to check if the issue persists.\n\nPlease let us know if there are further questions, thank you!", "votes": null }, { "id": "2848206", "postDate": "05/31/2024 21:59:15", "content": "

Hello, may I inquire about the publication of the scores and rankings after the removal of positions 12 and 15 in the first stage?

", "rawMarkdown": "Hello, may I inquire about the publication of the scores and rankings after the removal of positions 12 and 15 in the first stage?", "votes": null }, { "id": "2850161", "postDate": "06/02/2024 02:08:17", "content": "

Yes. We will post that tomorrow (June 2nd Noon EST)

", "rawMarkdown": "Yes. We will post that tomorrow (June 2nd Noon EST)", "votes": null }, { "id": "2851220", "postDate": "06/02/2024 15:54:04", "content": "

Here is the current ranking:
\nTeam Submission Public Score Private Score Overall Score
\n0002 VolETA 38292520 - Pub_0_20978 Priv_0_14300.csv 0.096977964 0.143001282 0.109762219
\n0004 ININVIAUN 38264465 - Pub_0_08046 Priv_0_19371.csv 0.087947783 0.193716212 0.117327902
\n0001 VCE 38222705 - Pub_0_20783 Priv_0_13209.csv 0.126116609 0.132090616 0.127776055
\n0003 FoodRiddle 38274835 - Pub_0_26261 Priv_0_15911.csv 0.164448493 0.159110558 0.162965733

", "rawMarkdown": "Here is the current ranking:\nTeam\tSubmission\tPublic Score\tPrivate Score\tOverall Score\n0002 VolETA\t38292520 - Pub_0_20978 Priv_0_14300.csv\t0.096977964\t0.143001282\t0.109762219\n0004 ININVIAUN\t38264465 - Pub_0_08046 Priv_0_19371.csv\t0.087947783\t0.193716212\t0.117327902\n0001 VCE\t38222705 - Pub_0_20783 Priv_0_13209.csv\t0.126116609\t0.132090616\t0.127776055\n0003 FoodRiddle\t38274835 - Pub_0_26261 Priv_0_15911.csv\t0.164448493\t0.159110558\t0.162965733", "votes": null }, { "id": "2851595", "postDate": "06/02/2024 20:24:01", "content": "

Thank you for your explanation and suggestion. We are looking forward to them.

\n

I just want to confirm that the final submission for Transformations is on June 4th and the date has not been extended, right? Also, do we need to submit them via email, or can we submit them here on Kaggle?

", "rawMarkdown": "Thank you for your explanation and suggestion. We are looking forward to them.\n\nI just want to confirm that the final submission for Transformations is on June 4th and the date has not been extended, right? Also, do we need to submit them via email, or can we submit them here on Kaggle?", "votes": null }, { "id": "2852670", "postDate": "06/03/2024 12:05:24", "content": "

Dear Organizers,

\n

We sent you an email with the Phase 2 submission data; we would love to check if you received our email titled \"Physically Informed 3D Food Reconstruction [MTF Challenge].\"

\n

Thanks a million!

", "rawMarkdown": "Dear Organizers,\n\nWe sent you an email with the Phase 2 submission data; we would love to check if you received our email titled \"Physically Informed 3D Food Reconstruction [MTF Challenge].\"\n\nThanks a million!", "votes": null } ], "comments": [ { "id": 2847926, "author_name": "umairharoon", "author_url": "", "post_date": "05/31/2024 18:08:29", "content": "

Dear Organizers,

\n

We are glad that our suggestion has been accepted. Yet, we found some issues with the new meshes:

\n\n
Mesh is not , no information on volume, barycenter and inertia tensor.\n
\n\n
RuntimeError: [Open3D Error] (std::shared_ptr<open3d::geometry::PointCloud> open3d::geometry::TriangleMesh::SamplePointsPoissonDisk(size_t, double, std::shared_ptr<open3d::geometry::PointCloud>, bool)) /root/Open3D/cpp/open3d/geometry/TriangleMesh.cpp:527: Input mesh has no triangles.\n
\n

", "votes": null, "replies": [ { "id": 2848029, "author_name": "metafoodcvpr", "author_url": "", "post_date": "05/31/2024 18:51:50", "content": "

Thanks for pointing these out!

\n
    \n
  1. While some of the meshes may not be watertight, we are still able to obtain the volume of these meshes and that is what is used in our calculation. We use trimesh python library for this volume calculation. Additionally, you can use the 3D print toolbox add-on in Blender and obtain the volume of the mesh too.
    2. \n
  2. Further, we are able to open the 6.obj file on all of our software and libraries including Windows 3D Viewer, Blender, MeshLab, and Trimesh. We haven't checked with Open3D but the format of the vertices, faces, normals and other information in the obj file is consistent with the rest so you shouldn't be facing any issues there. A quick suggestion could be redownloading the files and trying to open them without any modifications for sanity check. If it still fails to open, you can try other software or libraries to check if the issue persists.
    3. \n
\n

Please let us know if there are further questions, thank you!

", "votes": null, "replies": [ { "id": 2851595, "author_name": "umairharoon", "author_url": "", "post_date": "06/02/2024 20:24:01", "content": "

Thank you for your explanation and suggestion. We are looking forward to them.

\n

I just want to confirm that the final submission for Transformations is on June 4th and the date has not been extended, right? Also, do we need to submit them via email, or can we submit them here on Kaggle?

", "votes": null, "replies": [ { "id": 2852670, "author_name": "umairharoon", "author_url": "", "post_date": "06/03/2024 12:05:24", "content": "

Dear Organizers,

\n

We sent you an email with the Phase 2 submission data; we would love to check if you received our email titled \"Physically Informed 3D Food Reconstruction [MTF Challenge].\"

\n

Thanks a million!

", "votes": null, "replies": [] } ] } ] } ] }, { "id": 2848206, "author_name": "youngyang123", "author_url": "", "post_date": "05/31/2024 21:59:15", "content": "

Hello, may I inquire about the publication of the scores and rankings after the removal of positions 12 and 15 in the first stage?

", "votes": null, "replies": [ { "id": 2850161, "author_name": "metafoodcvpr", "author_url": "", "post_date": "06/02/2024 02:08:17", "content": "

Yes. We will post that tomorrow (June 2nd Noon EST)

", "votes": null, "replies": [] }, { "id": 2851220, "author_name": "metafoodcvpr", "author_url": "", "post_date": "06/02/2024 15:54:04", "content": "

Here is the current ranking:
\nTeam Submission Public Score Private Score Overall Score
\n0002 VolETA 38292520 - Pub_0_20978 Priv_0_14300.csv 0.096977964 0.143001282 0.109762219
\n0004 ININVIAUN 38264465 - Pub_0_08046 Priv_0_19371.csv 0.087947783 0.193716212 0.117327902
\n0001 VCE 38222705 - Pub_0_20783 Priv_0_13209.csv 0.126116609 0.132090616 0.127776055
\n0003 FoodRiddle 38274835 - Pub_0_26261 Priv_0_15911.csv 0.164448493 0.159110558 0.162965733

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "2846225": "Dear Phase 2 Participants,\n\nBoth Team VoIETA and VCE have raised issues with our dataset regarding incorrect meshes, scaling issues, and code related to point cloud transformations.\n\n1. **Incorrect Meshes for Scene 12 (steak) and Scene 15 (Chicken nugget):** We have traced back to our large dataset but were unable to find the original mesh scans for these two scenes. Therefore, we regret to inform you that we must remove these two scenes from all score calculations. We apologize for the inconvenience this may cause and appreciate your understanding.\n\n2. **Scaling Issues:** The first version of Phase 2 meshes we uploaded were not correctly scaled. We have re-uploaded the ground truth meshes in the `ground_truth_scaled` folder. Please use these for your final transformation calculations.\n\n3. **Point Cloud Transform Discrepancy:** Team VoIETA found a discrepancy in the point clouds after applying the transform in our script. However, our tests did not reveal this issue. Team VoIETA has provided a code snippet for their transformation function using Open3D. We will adapt their code and evaluate the results based on both our original transformation function and the Open3D one provided by Team VoIETA. We will use the minimum distance from the two as the final result. See the Phase 2 Data Are Now Available post for more details. \n\nThank you for your understanding and participation in the challenge. We apologize for any inconvenience caused.\n\nSincerely,\nMetaFood Workshop Organizing Committee", "2847926": "Dear Organizers, \n\nWe are glad that our suggestion has been accepted. Yet, we found some issues with the new meshes:\n\n* Some meshes are not watertight (e.g., mesh number 1), so we cannot see the actual volumes to validate phase 1. Example error from MeshLab:\n```bash\nMesh is not 'watertight', no information on volume, barycenter and inertia tensor.\n```\n* Mesh number 6 has no triangles, which is not parsable like other meshes. Example error:\n```bash\nRuntimeError: [Open3D Error] (std::shared_ptr open3d::geometry::TriangleMesh::SamplePointsPoissonDisk(size_t, double, std::shared_ptr, bool)) /root/Open3D/cpp/open3d/geometry/TriangleMesh.cpp:527: Input mesh has no triangles.\n```\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8304568%2Ffc360ec7775552006f7542042b452f45%2FScreenshot%20from%202024-05-31%2020-07-28.png?generation=1717178865781436&alt=media)", "2848029": "Thanks for pointing these out! \n1. While some of the meshes may not be watertight, we are still able to obtain the volume of these meshes and that is what is used in our calculation. We use trimesh python library for this volume calculation. Additionally, you can use the 3D print toolbox add-on in Blender and obtain the volume of the mesh too.\n2. Further, we are able to open the 6.obj file on all of our software and libraries including Windows 3D Viewer, Blender, MeshLab, and Trimesh. We haven't checked with Open3D but the format of the vertices, faces, normals and other information in the obj file is consistent with the rest so you shouldn't be facing any issues there. A quick suggestion could be redownloading the files and trying to open them without any modifications for sanity check. If it still fails to open, you can try other software or libraries to check if the issue persists.\n\nPlease let us know if there are further questions, thank you!", "2848206": "Hello, may I inquire about the publication of the scores and rankings after the removal of positions 12 and 15 in the first stage?", "2850161": "Yes. We will post that tomorrow (June 2nd Noon EST)", "2851220": "Here is the current ranking:\nTeam\tSubmission\tPublic Score\tPrivate Score\tOverall Score\n0002 VolETA\t38292520 - Pub_0_20978 Priv_0_14300.csv\t0.096977964\t0.143001282\t0.109762219\n0004 ININVIAUN\t38264465 - Pub_0_08046 Priv_0_19371.csv\t0.087947783\t0.193716212\t0.117327902\n0001 VCE\t38222705 - Pub_0_20783 Priv_0_13209.csv\t0.126116609\t0.132090616\t0.127776055\n0003 FoodRiddle\t38274835 - Pub_0_26261 Priv_0_15911.csv\t0.164448493\t0.159110558\t0.162965733", "2851595": "Thank you for your explanation and suggestion. We are looking forward to them.\n\nI just want to confirm that the final submission for Transformations is on June 4th and the date has not been extended, right? Also, do we need to submit them via email, or can we submit them here on Kaggle?", "2852670": "Dear Organizers,\n\nWe sent you an email with the Phase 2 submission data; we would love to check if you received our email titled \"Physically Informed 3D Food Reconstruction [MTF Challenge].\"\n\nThanks a million!" }, "source": "meta" }