YOLO TensorFlow
TensorFlow implementation of YOLO for real-time object detection using pretrained YOLO_small, YOLO_tiny, and YOLO_face models.
What is YOLO TensorFlow?
YOLO_tensorflow is a TensorFlow implementation of the YOLO (You Only Look Once) real-time object detection algorithm. It allows developers to perform fast object detection on images using pretrained YOLO models converted from Darknet. The project focuses on inference only, providing an easy way to integrate YOLO into Python-based computer vision applications without training overhead.
Developers and researchers working on computer vision projects who want to use YOLO for real-time object detection within the TensorFlow ecosystem, especially those needing a simple inference pipeline.
It offers a clean, inference-only TensorFlow port of YOLO with support for multiple pretrained models, making it easier to deploy YOLO in Python environments compared to the original C-based Darknet implementation.
Overview
tensorflow implementation of 'YOLO : Real-Time Object Detection'
Use Cases
Best For
- Adding real-time object detection to Python applications
- Experimenting with YOLO models in TensorFlow without training
- Quick prototyping of computer vision projects with pretrained detectors
- Converting Darknet YOLO weights to TensorFlow format
- Detecting objects in images with customizable output formats
- Integrating YOLO into workflows using OpenCV and TensorFlow
Not Ideal For
- Projects requiring training or fine-tuning of YOLO models on custom datasets
- Applications needing state-of-the-art object detection with recent YOLO variants like YOLOv3 or YOLOv4
- Commercial deployments due to the explicit non-commercial license restriction
- Environments with strict dependency management for newer TensorFlow versions beyond 2017 standards
Pros & Cons
Pros
Provides a clean API with methods like detect_from_file and detect_from_cvmat, allowing quick object detection from images or OpenCV matrices with minimal configuration.
Includes a weight converter tool to transform Darknet .weight files into TensorFlow checkpoint format, simplifying the process of porting pretrained YOLO models.
Supports console logging, visual display with bounding boxes, and export to image or text files, enabling versatile result handling for different workflows.
Offers YOLO_small, YOLO_tiny, and YOLO_face models, giving users options for balancing speed and accuracy without additional training.
Cons
The README explicitly states it does not support training, forcing users to rely on external tools like Darknet for model adaptation or fine-tuning.
Last updated in 2017, making it incompatible with modern TensorFlow releases and lacking bug fixes, performance improvements, or support for newer hardware.
Limited to older YOLO variants; misses contemporary architectures that offer better accuracy and features, such as YOLOv3 or YOLOv4, which are standard in current projects.
Frequently Asked Questions
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.