Post sammed3d - #249
Conversation
There was a problem hiding this comment.
Pull request overview
Adds a new review post to the site’s collections/_posts for the June 18 entry (SAM-Med3D). The PR also includes a substantial Gemfile.lock change, which appears inconsistent with the PR description (“Pas tenir compte de gemfile.lock”) and may be accidental.
Changes:
- Add new post
2026-06-18-SAMMed3D.mdwith images and references for the SAM-Med3D paper review. - Update
Gemfile.lockwith a different dependency resolution (including platform-specific entries).
Reviewed changes
Copilot reviewed 1 out of 7 changed files in this pull request and generated 3 comments.
| File | Description |
|---|---|
Gemfile.lock |
Lockfile dependency resolution/platform entries changed (likely unintended per PR description). |
collections/_posts/2026-06-18-SAMMed3D.md |
New review post content for SAM-Med3D, including figures and references. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
|
|
||
| # Method | ||
|
|
||
| ## SAM-Med3D Archiecture |
| $$\rightarrow$$ **Qualitative comparison** | ||
| <div style="text-align:center"> | ||
| <img src="/collections/images/SAMMed3D/comp_sammethods_quali.jpg" width=1500></div> | ||
| <p style="text-align: center;font-style:italic">: Figure 2. Visualization of SAM, SAM-Med2D, and SAM-Med3D across diverse anatomical structures and modalities for 1 or 5 points.</p> |
| # SA-Med3D-140K Dataset | ||
| SA-Med3D-140K is one of the largest volumetric medical segmentation datasets available at the time of publication. It combines data from 70 public datasets and 24 private datasets, resulting in 22k 3D medical images and 143k segmentation masks. | ||
|
|
||
| The dataset covers 28 imaging modalities, including CT, ultrasound, and 26 MRI sequences, spanning six major anatomical regions. It is split into 131k training masks and 12k validation masks. The training set contains 58 public datasets and all private data, while the validation set consists of 16 public benchmark datasets, some of which have their training set included the training sources. |
@JeanneRbs Au moment de merger, on ne peut pas sélectionner seulement une partie des changements. On accepte la PR en totalité, ou pas du tout. Il faut donc corriger le fichier sur ta branche avant de clôre la PR. Dans ton cas, comme tu as commité des changements sur Autrement, tu pourras répondre à la review automatique de Codex sur les typos et style dans le texte. |
Post du 18 juin,
Pas tenir compte de gemfile.lock