Skip to content

Latest commit

 

History

History
128 lines (94 loc) · 2.61 KB

map.md

File metadata and controls

128 lines (94 loc) · 2.61 KB
Raw

Map Training & Evaluation

Whenever training and evaluating, please edit the config paths (and checkpoint paths if testing) in the bash file. Also, change the data paths in the config files to your nuscenes raw data and processed annotation data.

During Evaluation, make sure to leave at least 1TB of storage space to store the BEV features. It would be great for subsequent work to apply this framework in an end-to-end manner to avoid storaging these BEV features.

One potential way would be training online mapping as an auxilliary task. During inference, only utilize the BEV features for trajectory prediction in an end-to-end manner.

MapTR

Training

Run

cd MapTR_modified/
source train.sh

Or by running:

export PYTHONPATH="${PYTHONPATH}:/MapBEVPrediction/MapTR_modified"

python tools/train.py \
  projects/configs/maptr/maptr_tiny_r50_24e.py \
  --deterministic \
  --no-validate

Evaluation

Run

cd MapTR_modified/
source test.sh

Or by running:

export PYTHONPATH="${PYTHONPATH}:/MapBEVPrediction/MapTR_modified"

python tools/test.py \
  projects/configs/maptr/maptr_tiny_r50_24e.py \
  work_dirs/maptr_tiny_r50_24e/YOURCHECKPOINT.pth \
  --eval chamfer \
  --bev_path /path_to_save_bev_features

MapTRv2

Training

Run

cd MapTRv2_modified/
source train.sh

Or by running:

export PYTHONPATH="${PYTHONPATH}:/MapBEVPrediction/MapTRv2_modified"

python tools/train.py \
  projects/configs/maptrv2/maptrv2_nusc_r50_24ep.py \
  --deterministic \
  --no-validate

Evaluation

Run

cd MapTRv2_modified/
source test.sh

Or by running:

export PYTHONPATH="${PYTHONPATH}:/MapBEVPrediction/MapTRv2_modified"

python tools/test.py \
  projects/configs/maptrv2/maptrv2_nusc_r50_24ep.py \
  work_dirs/maptrv2_nusc_r50_24ep/YOURCHECKPOINT.pth \
  --eval chamfer \
  --bev_path /path_to_save_bev_features

StreamMapNet

Training

Run

cd StreamMapNet_modified/
source train.sh

Or by running:

export PYTHONPATH="${PYTHONPATH}:/MapBEVPrediction/StreamMapNet_modified"

python tools/train.py \
  plugin/configs/nusc_newsplit_480_60x30_24e.py \
  --deterministic \
  --no-validate

Evaluation

Run

cd StreamMapNet_modified/
source test.sh

Or by running:

export PYTHONPATH="${PYTHONPATH}:/MapBEVPrediction/StreamMapNet_modified"

python tools/test.py \
  plugin/configs/nusc_newsplit_480_60x30_24e.py \
  work_dirs/nusc_newsplit_480_60x30_24e/YOURCHECKPOINT.pth \
  --eval \
  --bev_path /path_to_save_bev_features