New process for poseEstimation using multiple processes

.github/workflows/continuous-integration.yml

@@ -25,6 +25,8 @@ jobs:
         os: [ubuntu-latest, windows-latest, macos-latest, macos-13] # Mac ARM M1 vs Intel processors
         python-version: ["3.10", "3.11"]
 
+    timeout-minutes: 720
+
     steps:
       - name: Checkout code
         uses: actions/checkout@v4

Pose2Sim/Demo_Batch/Config.toml

@@ -34,7 +34,16 @@ exclude_from_batch = [] # List of trials to be excluded from batch analysis, ['<
 
 
 [pose]
-vid_img_extension = 'mp4' # any video or image extension
+rotation = 0  # Degree of rotation applied to the image, allowed in multiples of 90. Use [90, 90, ...] for different angles per source.
+vid_img_extension = 'mp4'  # File extension for videos or images, or 'webcam' for webcam input.
+
+webcam_ids = 0  # Webcam ID to use. Default is 0. For multiple webcams, use a list like [0, 1, ...].
+capture_mode = 'continuous'  # Only useful for webcams. 'Continuous' captures frames as soon as buffer space is available; 'alternating' waits until the buffer is free.
+webcam_recording = false  # If True, saves webcam streams while analyzing them to prevent frame loss due to processing delays.
+
+combined_frames = true  # Accelerates processing by batch processing frames. (Beta feature)
+multi_workers = false  # Enables multiple worker processes for parallel frame analysis. (Not recommended)
+max_simultaneous_sources = 9  # Limits the number of sources processed simultaneously. Recommended values: 4, 9, 16, ... (perfect square).
 
 pose_model = 'Body_with_feet'  #With RTMLib: 
                          # - Body_with_feet (default HALPE_26 model), 

Pose2Sim/Demo_MultiPerson/Config.toml

@@ -34,7 +34,16 @@ exclude_from_batch = [] # List of trials to be excluded from batch analysis, ['<
 
 
 [pose]
-vid_img_extension = 'mp4' # any video or image extension
+rotation = 0  # Degree of rotation applied to the image, allowed in multiples of 90. Use [90, 90, ...] for different angles per source.
+vid_img_extension = 'mp4'  # File extension for videos or images, or 'webcam' for webcam input.
+
+webcam_ids = 0  # Webcam ID to use. Default is 0. For multiple webcams, use a list like [0, 1, ...].
+capture_mode = 'continuous'  # Only useful for webcams. 'Continuous' captures frames as soon as buffer space is available; 'alternating' waits until the buffer is free.
+webcam_recording = false  # If True, saves webcam streams while analyzing them to prevent frame loss due to processing delays.
+
+combined_frames = true  # Accelerates processing by batch processing frames. (Beta feature)
+multi_workers = false  # Enables multiple worker processes for parallel frame analysis. (Not recommended)
+max_simultaneous_sources = 9  # Limits the number of sources processed simultaneously. Recommended values: 4, 9, 16, ... (perfect square).
 
 pose_model = 'Body_with_feet'  #With RTMLib: 
                          # - Body_with_feet (default HALPE_26 model), 

Pose2Sim/Demo_SinglePerson/Config.toml

@@ -34,7 +34,16 @@ exclude_from_batch = [] # List of trials to be excluded from batch analysis, ['<
 
 
 [pose]
-vid_img_extension = 'mp4' # any video or image extension
+rotation = 0  # Degree of rotation applied to the image, allowed in multiples of 90. Use [90, 90, ...] for different angles per source.
+vid_img_extension = 'mp4'  # File extension for videos or images, or 'webcam' for webcam input.
+
+webcam_ids = 0  # Webcam ID to use. Default is 0. For multiple webcams, use a list like [0, 1, ...].
+capture_mode = 'continuous'  # Only useful for webcams. 'Continuous' captures frames as soon as buffer space is available; 'alternating' waits until the buffer is free.
+webcam_recording = false  # If True, saves webcam streams while analyzing them to prevent frame loss due to processing delays.
+
+combined_frames = true  # Accelerates processing by batch processing frames. (Beta feature)
+multi_workers = false  # Enables multiple worker processes for parallel frame analysis. (Not recommended)
+max_simultaneous_sources = 9  # Limits the number of sources processed simultaneously. Recommended values: 4, 9, 16, ... (perfect square).
 
 pose_model = 'Body_with_feet'  #With RTMLib: 
                          # - Body_with_feet (default HALPE_26 model), 
@@ -76,7 +85,7 @@ mode = 'balanced' # 'lightweight', 'balanced', 'performance',
 #          'pose_input_size':[256,256]}"""
 
 
-det_frequency = 4 # Run person detection only every N frames, and inbetween track previously detected bounding boxes (pose estimation is still run on all frames). 
+det_frequency = 1 # Run person detection only every N frames, and inbetween track previously detected bounding boxes (pose estimation is still run on all frames). 
                   # Equal to or greater than 1, can be as high as you want in simple uncrowded cases. Much faster, but might be less accurate. 
 device = 'auto' # 'auto', 'CPU', 'CUDA', 'MPS', 'ROCM'
 backend = 'auto' # 'auto', 'openvino', 'onnxruntime', 'opencv'