Feature/detect format function (#144)

* CDL: minor doc typo fix

* Undoing some changes that got mixed in

* Add detect_pose_format function and SupportedPoseFormat Literal

* detect_known_pose_format and tests for it.

* various cleanup changes, style changes

* missing import

* undo black formatting for face contours and ignore_names

* SupportedPoseFormat->KnownPoseFormat

* Unreachable raise ValueErrors fixed

* generic utils type annotations

* change detect_known_format to take Pose or PoseHeader

* Reraise ImportError if mediapipe is not installed

* conftest update to supply unknown-format fake poses

* nicer formatting for plane_info and line_info

* fix import in generic_test.py

* add some pylint disables, consistent with pose-evaluation

* Change import in conftest.py

* change import style in generic.py

* change more imports

* Fix a few type issues

* Change matrix strategy fail-fast to false, so that we can still run tests if Python 3.8 does not work

* Union for type annotation backwards compatibility

* Add checks for NotImplementedError

* Fix correct_wrist modifying input, and wrong shape for stacked conf. Also added a function to check fake_pose and its outputs

* Simplify get_component_names and fix spacing

* fix test_get_component_names

Author: cleong110

.github/workflows/python.yaml

@@ -13,6 +13,7 @@ jobs:
     strategy:
       matrix:
         python-version: ["3.8", "3.9", "3.10", "3.11", "3.12"]
+      fail-fast: false
 
     steps:
       - uses: actions/checkout@v3

.gitignore

@@ -1,2 +1,6 @@
 .idea/
 .DS_Store
+.vscode/
+.coverage
+.coveragerc
+coverage.lcov

src/python/ComfyUI-Pose-Format/nodes.py

@@ -2,7 +2,7 @@
 
 import cv2
 import torch
-from pose_format import Pose
+from pose_format.pose import Pose
 from pose_format.pose_visualizer import PoseVisualizer
 from pose_format.utils.generic import reduce_holistic
 from pose_format.utils.openpose import OpenPose_Components

src/python/pose_format/bin/pose_visualizer.py

@@ -3,7 +3,7 @@
 import argparse
 import os
 
-from pose_format import Pose
+from pose_format.pose import Pose
 from pose_format.pose_visualizer import PoseVisualizer
 from pose_format.utils.generic import pose_normalization_info
 

src/python/pose_format/pose.py

@@ -1,5 +1,5 @@
 from itertools import chain
-from typing import BinaryIO, Dict, List, Tuple, Type
+from typing import BinaryIO, Dict, List, Tuple, Type, Union
 
 import numpy as np
 import numpy.ma as ma
@@ -87,7 +87,7 @@ def focus(self):
         dimensions = (maxs - mins).tolist()
         self.header.dimensions = PoseHeaderDimensions(*dimensions)
 
-    def normalize(self, info: PoseNormalizationInfo|None=None, scale_factor: float = 1) -> "Pose":
+    def normalize(self, info: Union[PoseNormalizationInfo,None]=None, scale_factor: float = 1) -> "Pose":
         """
         Normalize the points to a fixed distance between two particular points.
 
@@ -203,7 +203,7 @@ def frame_dropout_normal(self, dropout_mean: float = 0.5, dropout_std: float = 0
         body, selected_indexes = self.body.frame_dropout_normal(dropout_mean=dropout_mean, dropout_std=dropout_std)
         return Pose(header=self.header, body=body), selected_indexes
 
-    def get_components(self, components: List[str], points: Dict[str, List[str]] = None):
+    def get_components(self, components: List[str], points: Union[Dict[str, List[str]],None] = None):
         """
         get pose components based on criteria.
 

src/python/pose_format/utils/conftest.py

@@ -0,0 +1,23 @@
+import copy
+from typing import List, get_args
+import pytest
+from pose_format.pose import Pose
+from pose_format.utils.generic import get_standard_components_for_known_format, fake_pose, KnownPoseFormat
+
+@pytest.fixture
+def fake_poses(request) -> List[Pose]:
+    # Access the parameter passed to the fixture
+    known_format = request.param
+    count = getattr(request, "count", 3)  
+    known_formats = get_args(KnownPoseFormat)
+    if known_format in known_formats:
+
+        components = get_standard_components_for_known_format(known_format)
+        return copy.deepcopy([fake_pose(i * 10 + 10, components=components) for i in range(count)])
+    else:
+        # get openpose
+        fake_poses_list = [fake_pose(i * 10 + 10) for i in range(count)]
+        for i, pose in enumerate(fake_poses_list):
+            for component in pose.header.components:
+                component.name = f"unknown_component_{i}_formerly_{component.name}"
+        return copy.deepcopy(fake_poses_list)