/image-processing

Small tool to segment or detect objects in images with tensorflow

Primary LanguagePython

Image processing

Segment or detect objects on images using tensorflow. Models are from different sources

Usage

launcher.py [-h] -t MODEL_TYPE [-vf VIDEO_FILE] [-di IN_DIRECTORY] -do OUT_DIRECTORY

Videos support

This tool supports video using the -vf flag. Input videos can be in any format supported by OpenCV. The output video is written using the MJPEG codec.

Examples

Segmentation using deeplabv3 from tf research repo.

python3 launcher.py -di /dir/with/images -do /output/directory -t segmentationcs

Note:

To use this model, you need to download one pretrained model from [https://github.com/tensorflow/models/blob/master/research/deeplab/g3doc/model_zoo.md].

Tested only with deeplabv3_mnv2_cityscapes_train_2018_02_05.tar.gz

Segmentation

python3 launcher.py -di /dir/with/images -do /output/directory -t segmentation

Detection

python3 launcher.py -di /dir/with/images -do /output/directory -t detection