The code in this repository provides an implementation for the methods discussed in the following paper: Fully Convolutional One-Shot Object Segmentation for Industrial Robotics. The paper is dedicated to the memory of our wonderful colleague and friend Benjamin Schnieders, a bright young scientist, who recently passed away.
To run the program using the ObjectDetection@Work dataset, a preprocessed dataset can be obtained from:
Filesize: 11.9 GB
Upon downloading an extracting the dataset, the framework can be run directly via:
python codetect_obj.py --logProgress --holdout1=0 --holdout2=1
where holdout1 and holdout2 represent the indexes belonging to the holdout objects.
Alternatively, the following script can be run to test each holdout combination: