Update modelzoo documentation fmpose3d

deruyter92 · Mar 20, 2026 · Feb 25, 2026 · Feb 25, 2026 · Feb 25, 2026 · Feb 25, 2026
commit de1c5735b129199b5252a88acca3c597b31ea4fc
diff --git a/docs/ModelZoo.md b/docs/ModelZoo.md
@@ -95,20 +95,26 @@ deeplabcut.video_inference_superanimal([video_path],
 ### Practical example: FMPose3D monocular 3D inference
 
 For FMPose3D models, use `model_name="fmpose3d_animals"` or
-`model_name="fmpose3d_humans"`. In this branch, `superanimal_name` is
-accepted for API compatibility but ignored for model selection.
+`model_name="fmpose3d_humans"`. The parameter `superanimal_name` is
+accepted for API compatibility but ignored for model selection. 
+
+Like the 2D superanimal models, this inference branch writes
+intermediate 2D predictions to `<video>_DLC_fmpose3d_*.h5` and `<video>_DLC_fmpose3d_*.json`. In addition, 3D predictions are saved to
+`<video>_DLC_fmpose3d_*_3d.h5` and `<video>_DLC_fmpose3d_*_3d.json`. Set `fmpose_return_3d=True` to also return
+the in-memory 3D dataframe (`df_3d`) in the function output.
 
 ```python
 import deeplabcut
 
 video_path = "demo-video.mp4"
-deeplabcut.video_inference_superanimal(
+result = deeplabcut.video_inference_superanimal(
     videos=[video_path],
-    superanimal_name="superanimal_quadruped",  # ignored for fmpose3d models
+    superanimal_name="",  # ignored for fmpose3d models
     model_name="fmpose3d_animals",
     batch_size=8,
     fmpose_return_3d=True,  # include 3D dataframe in returned payload
 )
+df_3d = result[video_path]["df_3d"]
 ```