https://github.com/kienpc1234/kientf_object_detection
KienTF Object Detection is an improved version of the original object_detection repository, addressing various issues and enhancing stability. This project aims to fix bugs, optimize performance, and ensure better compatibility for real-world object detection tasks.
https://github.com/kienpc1234/kientf_object_detection
kientf machine-learning object-detection opencv pakage pip python python3 tensorflow tensorflow-lite tensorflow2
Last synced: 2 months ago
JSON representation
KienTF Object Detection is an improved version of the original object_detection repository, addressing various issues and enhancing stability. This project aims to fix bugs, optimize performance, and ensure better compatibility for real-world object detection tasks.
- Host: GitHub
- URL: https://github.com/kienpc1234/kientf_object_detection
- Owner: KienPC1234
- License: apache-2.0
- Created: 2025-02-17T09:00:55.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-04-15T15:14:31.000Z (about 1 year ago)
- Last Synced: 2025-12-27T01:36:10.247Z (6 months ago)
- Topics: kientf, machine-learning, object-detection, opencv, pakage, pip, python, python3, tensorflow, tensorflow-lite, tensorflow2
- Language: Python
- Homepage:
- Size: 19.7 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# 🎨 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**:
```bash
pip install git+https://github.com/KienPC1234/kientf_object_detection.git
```
### 🔹 Dependencies
Ensure you have **TensorFlow 2.x** installed (version **<2.12**):
```bash
pip install "tensorflow<2.12"
```
---
## 🔧 Usage
### 🔹 1. Train a Model
Modify the pipeline config file and run:
```bash
python model_main_tf2.py --pipeline_config_path=configs/my_model.config --model_dir=training/
```
### 🔹 2. Export Trained Model
```bash
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
```python
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](LICENSE) for details.
---
## 📩 Contact
For any questions or collaborations, reach out via:
📧 **Email:** kienpc872009@gmail.com
🔥 **Happy coding & object detecting!** 🔥