How shp50c/sign_language_using_object_detection_yolo_v5 works
Project type and purpose
• This is a modular Python‐based computer vision application with two main functions:
a) A machine‐learning training pipeline that collects, validates, trains (using YOLOv5), and pushes models.
b) A Flask‐based web interface for inference (live video or image prediction).
• It bundles the official YOLOv5 code (in the yolov5/ folder), custom pipeline modules (src/), data collection (data_collector.py), and a minimal front end (templates/index.html).
File structure
• Top‐level artifacts:
• data_collector.py – script to gather images.
• app.py – Flask application for serving inference.
• Dockerfile / .dockerignore – containerization.
• template.py / setup.py / requirements.txt – project setup and scaffolding.
• src/ – core pipeline modules segmented by functionality:
• components/ – data ingestion, validation, model training, model pushing.
• configuration/ – S3 interactions.
• constant/ – application and pipeline constants.
• entity/ – data structures for configs and artifacts.
• pipeline/ – orchestrates the end‐to‐end training pipeline.
• logger.py / exception.py / utils.py – cross‐cutting concerns.
• templates/ – HTML UI.
• yolov5/ – third‐party detection/segmentation code, model definitions, trainers, predictors, utilities, Docker support for serving.
README for additional insights
• Confirms:
• Seven‐class sign language dataset augmented to 8,400 images.
• Modular pipeline mirrors a Jupyter notebook workflow.
• Logs under logs/ for tracing pipeline steps and inference events.
Elements
• Labels: name each module (“DataIngestion”, “ModelTrainer”, “FlaskAPI”, etc.).
• Gray docker boundary to show container limits