kitti.md

KITTI Datasets

We test on the KITTI RAW sequences

Exporting Sequence Poses for pythonic usage

Please refer to this naive method to produce poses for every sequence.

Dataset validation / visualization in ROS (Optional)

Check kitti_visualize repo.

Training Schedule

Baseline:

## copy example config
cd config
cp kitti_wpose_example kitti_wpose.py

## Modify config path
nano kitti_wpose.py
cd ..

## Train
./launcher/train.sh configs/kitti_wpose.py 0 $experiment_name

## Evaluation
python3 scripts/test.py configs/kitti_wpose.py 0 $CHECKPOINT_PATH

It's fine to just use the baseline model for projects. After training baseline, you can further re-train with self-distillation:

## export checkpoint
python3 monodepth/transform_teacher.py $Pretrained_checkpoint $output_compressed_checkpoint

## copy example config 
cd config
cp distill_kitti_example distill_kitti.py

## Modify config path and checkpoint path based on  $output_compressed_checkpoint
nano distill_kitti.py
cd ..

## Train
./launcher/train.sh configs/distill_kitti.py 0 $experiment_name

Visualize with jupyter notebook

Check demos/demo.ipynb for visualizing datasets and simple demos.

Onnx export

We support exporting pretrained model to onnx model, and you need to install onnx and onnxruntime.

python3 scripts/onnx_export.py $CONFIG_FILE $CHECKPOINT_PATH $ONNX_PATH 

Online ROS full demo

1. Launch kitti_visualize to stream image data topics and Rviz visualization.
2. Launch monodepth_ros to infer on camera topics.