This is an object recognition system designed as a component of the ARGUS project. It uses TensorFlow for real-time object detection via your computer’s webcam. The system is capable of identifying a variety of objects and displaying bounding boxes around them along with confidence scores.
Ensure the following dependencies are installed:
• Python 3.7+
• TensorFlow 2.x
• OpenCV
• NumPy
You can install the dependencies using the following command:
pip3.10 install tensorflow opencv-python-headless numpy
1. Place your TensorFlow model in the specified directory: /Users/blakeweiss/Desktop/objectrecognitionARGUS/pretrainedmodelobj/. If your model is stored elsewhere, update the path in the script accordingly.
2. Connect a webcam to your computer for real-time object detection.
To run the object recognition system, execute the script:
python3.10 your_script_name.py
The program will capture video from the webcam. During this a window will be displayed as will the detected objects, their bounding boxes, and confidence levels.
Press q to exit the application.
• Confidence Threshold: Modify the detection confidence threshold by changing the confidence > 0.5 condition within the objectrecognitionrun function to a different value.
• Label Map: The label map dictionary links numerical class IDs to their respective object names. Adjust this dictionary as needed to align with the labels provided by your TensorFlow model.
• Ensure that the TensorFlow model is correctly loaded and the video feed is functional if no objects are detected.
• Verify that all dependencies are installed and properly configured.
• If the webcam does not work, try changing the device index in cv2.VideoCapture(0) to another index if multiple cameras are available.
This component is licensed under the MIT License.
Feel free to adjust this README to better fit your project’s needs or specific details.