Supported Modules#

The tables below show all the currently supported modules integrated in pyppbox in different ways. The main goal is to make them fully compatible with our PoseTReID framework read more here. Thus, not every original feature of each supported module is fully functioning as originally made. For example, the supported YOLO_Classic / YOLO module can only be used as a detector (Inference/Prediction) with official `.weights` files, but not for custom data training, etc.

Supported Detectors#
General Name	Config Name	Details
YOLO	YOLO_Classic	Built-in by using OpenCV DNN Model: .weights V2, V3, V4 Run on: CPU or GPU (OpenCV DNN)
YOLO	YOLO_Ultralytics	Integrated by linking ultralytics Model: .pt V3, V5, V8, V9, v10, v11, v12, v26 Run on: CPU or GPU (PyTorch)

GT is also a supported detector name. It reads detections from a ground-truth text file; see Example 10: Use GT (Ground-truth) as a detector. Set both tracker and reider to "None" to preserve the GT identities automatically. With other stages enabled, pyppbox.ppb.mt.MT.forceFullGTMode() overrides detection-only GT behavior until the next module selection.

Supported Trackers#
General Name	Config Name	Details
Centroid	Centroid	Built-in / Native Run on: CPU
SORT	SORT	Integrated by embedding SORT repo Run on: CPU
DeepSORT	DeepSORT	Integrated by embedding DeepSORT repo Appearance encoder: TensorFlow; assignment: CPU

Supported ReIDers#
General Name	Config Name	Details
FaceNet	FaceNet	Integrated by embedding FaceNet repo Run on: CPU or GPU (Tensorflow)
Torchreid	Torchreid	Integrated by linking pyppbox-torchreid Model: OSNet-AIN, OSNet, MLFN Run on: CPU or GPU (PyTorch)