This is an implementation for TensorFlow of the PoseNet architecture
As described in the ICCV 2015 paper PoseNet: A Convolutional Network for Real-Time 6-DOF Camera Relocalization Alex Kendall, Matthew Grimes and Roberto Cipolla [http://mi.eng.cam.ac.uk/projects/relocalisation/]
-
Download the Cambridge Landmarks King's College dataset from here.
-
Download the starting and trained weights from here.
-
The PoseNet model is defined in the posenet.py file
-
The starting and trained weights (posenet.npy and PoseNet.ckpt respectively) for training were obtained by converting caffemodel weights from here and then training.
-
To run:
- Extract the King's College dataset to wherever you prefer
- Extract the starting and trained weights to wherever you prefer
- Update the paths on line 13 (train.py) as well as lines 15 and 17 (test.py)
- If you want to retrain, simply run train.py (note this will take a long time)
- If you just want to test, simply run test.py