non-offical implementation of undeepvo
-
Prepare dataset run generate_train.m with matlab
-
Activate tensorflow source ~/tensorflow/bin/activate
-
Training python undeepvo_main.py --mode train --model_name my_model --data_path /media/youngji/storagedevice/naver_data/kitti_odometry/dataset/ --filenames_file ./utils/filenames/kitti_train_files2.txt --log_directory ./tmp/
-
visualization tensorboard --logdir=./tmp/my_model
-
Testing python monodepth_main.py --mode test --data_path /media/youngji/storagedevice/naver_data/kitti_odometry/dataset/
--filenames_file ./utils/filenames/kitti_train_files2.txt --log_directory ./tmp/
--checkpoint_path ./tmp/my_model/model-181250
python undeepvo_test.py --checkpoint_path ./tmp/my_model3/model-30000