imgalz: A Modular Library for Image Analysis

Installation

pip install -r requirements.txt
pip install -e .

Running imgalz

Predicting on Videos

Run main.py to test the tracker on all videos in the resources folder.

python tools/main.py cfg/video_track.py

If you want to run it on a webcam, use:

python tools/main.py cfg/video_track.py --cfg-options data.data_root=webcam

Predicting/Validating on Images

If you want to predict on images, use:

python tools/main.py cfg/imgdet.py

If you want to validate the model trained on COCO in the YOLO dataset, refer to the configuration in cfg/imgdet_val.py.

Validating on a Custom YOLO Dataset/Model

Define the class names using class_name=[0:'person',1:'car'....] and pass class_name to hook.detval.

Optional Models

Detectors

• YOLOv5
• YOLOv8
• YOLOv8pose(person)
• YOLOv8seg

Tracking

• ByteTrack
• Motpy
• NorFair
• OCSort

Weights

The ONNX model in the example is exported directly from the official code and can be obtained from the Google Drive.

Other Questions

1. Why did the bounding box (bbox) decoding fail?

   Please check the configuration file and ensure that pipeline.type is compatible with the model you intend to use.