fix: docs + add procedural staircase skips in AMASS

Author: RumblingTurtle

docs/downloading_data.md

@@ -4,7 +4,7 @@ Run ```./prepare_data.sh``` script to prepare the folder structure for the datas
 
 ## AMASS dataset
 
-1) Register on the [AMASS website](https://amass.is.tue.mpg.de/) and download the datasets you want to use. Make sure to download SMPL-X data (gender doesn't matter). Put the downloaded files into ```./assets/amass/datasets``` subfolder.
+1) Register on the [AMASS website](https://amass.is.tue.mpg.de/) and download the datasets you want to use. Make sure to download SMPL-H data (gender doesn't matter). Put the downloaded files into ```./assets/amass/datasets``` subfolder.
 
 2) Register and download the [SPML-H model](https://download.is.tue.mpg.de/download.php?domain=mano&resume=1&sfile=smplh.tar.xz) and extract it into ```./assets/amass/body_models``` subfolder
 

recap/amass/data_loader.py

@@ -19,9 +19,11 @@ def __init__(
         shuffle=False,
         name_blacklist=None,
         name_whitelist=None,
+        skip_stairs=False,
         device="cpu",
     ):
         self.datasets_path = datasets_path
+        self.skip_stairs = skip_stairs
         self.device = device
         self.target_fps = target_fps
         self.template_scale = torch.tensor(template_scale, device=device)
@@ -155,7 +157,11 @@ def preprocess(self, path) -> AMASSMotionWrapper:
             )
 
             heel_indices = [SMPLH_JOINT_NAMES_TO_IDX["left_ankle"], SMPLH_JOINT_NAMES_TO_IDX["right_ankle"]]
-            if torch.any(positions[:, heel_indices, 2] > positions[:, 0, 2].mean() * 0.3).sum().item() > 0:
+            if (
+                self.skip_stairs
+                and torch.any(positions[:, heel_indices, 2] > positions[:, 0, 2].mean() * 0.3).sum().item() > 0
+            ):
+                print("Stairs present in the clip")
                 return None
 
             return clip_name, AMASSMotionWrapper(positions=positions, rotations=rotations)
@@ -189,6 +195,7 @@ def _load_and_validate_clip(self, path):
                 return None
 
         if seq_length is None or seq_length < 10:
+            print("sequence duration is too small")
             return None
 
         # Trim sequences to valid length