KhachKhach 2.0.3

A package for processing video frames, annotating keypoints, and more.

KhachKhach ✂️

<<<<<<< HEAD KhachKhach is a Python library for advanced video and image processing, with a focus on frame extraction, keypoint annotation, and object detection using YOLO models. It is designed to make computer vision workflows easy, flexible, and highly customizable....

KhachKhach is a Python library for advanced video and image processing, with a focus on frame extraction, keypoint annotation, and object detection using YOLO models. It is designed to make computer vision workflows easy, flexible, and highly customizable.

ce34636f857fd468038edcffc16d3643792b6176

Unique Features

• Flexible YOLO Model Support: Use any Ultralytics YOLO model (including pose models) for detection and keypoint annotation.
• Frame Extraction: Extract frames from videos at custom intervals, time ranges, or frame counts.
• Keypoint Annotation: Annotate single images or entire folders with keypoints, supporting normalized coordinates and multiple output formats.
• Bounding Box Processing: Detect and annotate bounding boxes, with options for saving annotated images and exporting results.
• Batch Processing: Process entire folders of images or videos in one go.
• Custom Output: Choose between normalized (0-1) or absolute coordinates, and select output format (space-separated, comma-separated, etc).
• Easy Integration: Simple API for use in scripts, research, or production pipelines.
• Utilities: Includes tools for appending text to files, extracting XYN arrays, and more.

Installation

Install the required dependencies and KhachKhach via pip:

pip install opencv-python numpy pillow ultralytics
pip install KhachKhach

Quick Start

1. Extract Frames from a Video

import khachkhach as kk

video_path = "your_video.mp4"
frames_dir = "frames"

video_processor = kk.VideoProcessor()
video_processor.extract_frames(video_path, frames_dir)

2. Detect Objects or Annotate Keypoints

import khachkhach as kk

engine = kk.DetectionEngine("yolo11n-pose.pt")
engine.detect_objects(input_path="frames", output_dir="objectin")
# or for keypoints:
engine.annotate_keypoints(input_path="frames", output_dir="objectin", normalize_coords=True)

3. Full Pipeline Example

import os
import khachkhach as kk

video_path = "your_video.mp4"
frames_dir = "frames"
output_dir = "objectin"

os.makedirs(frames_dir, exist_ok=True)
os.makedirs(output_dir, exist_ok=True)

video_processor = kk.VideoProcessor()
video_processor.extract_frames(video_path, frames_dir)

engine = kk.DetectionEngine("yolo11n-pose.pt")
engine.annotate_keypoints(input_path=frames_dir, output_dir=output_dir)

Advanced Usage

• Normalized Coordinates: Get keypoints in 0-1 range for ML workflows.
• Custom Output Format: Choose between space-separated or comma-separated output.
• Batch Processing: Process all images in a folder with a single call.
• Integration: Use in your own scripts or extend with custom logic.

Example Scripts

See the test folder for ready-to-run examples:

• test_detect.py: Object detection on images.
• test_video.py: Full pipeline from video to keypoints.
• test.py: Advanced keypoint extraction with normalization.

Requirements

• Python 3.7+
• opencv-python
• numpy
• pillow
• ultralytics

License

See LICENSE for details.