🎨 KienTF Object Detection

A refined and bug-fixed version of TensorFlow's Object Detection API

---

📌 Overview

kientf_object_detection is an improved version of the original TensorFlow Object Detection API, focusing on:
✅ Fixing critical bugs that impact model training and inference.
✅ Enhancing performance and stability for real-world object detection tasks.
✅ Ensuring compatibility with TensorFlow 2.x (up to 2.12) for smoother workflows.

If you're tired of running into frustrating errors in the original object_detection, this repository provides a cleaner, more reliable solution!

---

🚀 Key Features & Fixes

🔧 Bug Fixes:

• Resolved major issues in model training, inference, and evaluation.
• Fixed deprecated functions and compatibility errors with TensorFlow 2.x.

⚡ Performance Enhancements:

• Optimized data processing & augmentation pipelines.
• Improved model stability and efficiency in large-scale datasets.

📊 Better Evaluation Metrics:

• Fixed inconsistencies in mAP calculations.
• More accurate and reliable benchmarking tools.

🔄 Easy Integration:

• Maintains full compatibility with existing TensorFlow Object Detection API models.
• Plug-and-play support for both custom datasets and pre-trained models.

---

👥 Installation

You can install this repository directly via pip:

pip install git+https://github.com/KienPC1234/kientf_object_detection.git

🔹 Dependencies

Ensure you have TensorFlow 2.x installed (version <2.12):

pip install "tensorflow<2.12"

---

🔧 Usage

🔹 1. Train a Model

Modify the pipeline config file and run:

python model_main_tf2.py --pipeline_config_path=configs/my_model.config --model_dir=training/

🔹 2. Export Trained Model

python exporter_main_v2.py --input_type image_tensor --pipeline_config_path=configs/my_model.config --trained_checkpoint_dir=training/ --output_directory=exported_model/

🔹 3. Run Inference

from object_detection.utils import visualization_utils as viz_utils
from object_detection.builders import model_builder

# Load the model and run inference

---

🤝 Contributing

Contributions are welcome! Feel free to open an issue or submit a pull request.

💡 Have a bug to report? Open an issue!
🚀 Want to improve the repo? Submit a PR!

---

📝 License

This project is licensed under the Apache 2.0 License. See LICENSE for details.

---

📩 Contact

For any questions or collaborations, reach out via:
📧 Email: kienpc872009@gmail.com

🔥 Happy coding & object detecting! 🔥