sfMLearning的prepare data调试结果
KITTI数据下载方法,上一个博客应该已经说过:
For downloading all raw data from the KITTI websites, create a new folder, copy this script into the folder and execute it from the command line:
./raw_data_downloader.sh
It will download the zip files and extract them into a coherent data structure: Each folder contains all sequences recorded at a single day, including the calibration files for that day.
Rawdata文件
终端命令行:
python data/prepare_train_data.py --dataset_dir=I:/raw_data/ --dataset_name="kitti_raw_eigen" --dump_root=I:/raw_data/formatted/data/ --seq_length=3 --img_width=416 --img_height=128 --num_threads=4
解释:(–dataset_dir待处理数据的文件路径,–dump_root处理好的文件输出到的文件保存路径)
python data/prepare_train_data.py --dataset_dir=/path/to/raw/kitti/dataset/ --dataset_name='kitti_raw_eigen' --dump_root=/path/to/resulting/formatted/data/ --seq_length=3 --img_width=416 --img_height=128 --num_threads=4
train终端命令行
python train.py --dataset_dir=I:/raw_data/formatted/data/ --checkpoint_dir=I:/raw_data/checkpoints/ --img_width=416 --img_height=128 --batch_size=4
解释:
python train.py --dataset_dir=/path/to/the/formatted/data/ --checkpoint_dir=/where/to/store/checkpoints/ --img_width=416 --img_height=128 --batch_size=4
查看运行状态:
tensorboard --logdir=I:/raw_data/checkpoints/ --port=8888
等待运行结束,大概需要几个小时,未完待续!!!