It is a graphical image annotation tool. Designed a python based auto labeling tool ,where in we can load a tensorflow model and images folder.Model runs inference on images and draws bounding box and class label.
- Clone this repository.
- Go to the directory where label.py exists.
- Download Faster R-CNN, Mask R-CNN, ssdlite_mobilenet_v2 models from this drive link.Unzip files and copy them in the directory.
- Label_Image
- faster_rcnn
- Mask_RCNN
- ssdlite_mobilenet_v2
- label.py
- label.txt
- detection.py ( detailed code to get detected labels and co ordinates from model )
- Run the python file.
python label.py
- Then there appears UI. Select images folder by clicking 'Directory' and then once loaded click 'Select Image' to select desired image.
- Select model and Detection threshold as required.
- Also select object that you want the model to detect.
- Click 'Detect' and Save annotations of detected objects.
- This tool confines to the models Faster R-CNN, Mask R-CNN, ssdlite_mobilenet_v2.
- Download tensorflow models from here.
- List model names in model_list array in label.py.
- Make sure you create a label.txt where in object names of your Dataset are listed.