modules/tools/amodel/README.md
amodel is Apollo's model deployment and management tool.
If you are running in Apollo docker, you can skip this step. If you are running outside of docker, the following environment needs to be set up.
source docker/setup_host/host_env.sh
amodel provides the following commands:
You can get the installed models in Apollo through the list command.
$ amodel list
Name |Task_type |Sensor_type |Framework |Date
mask_pillars |3d_detection |lidar |paddlepaddle |2021-07-30
center_point |3d_detection |lidar |paddlepaddle |2022-07-22
point_pillars |3d_detection |lidar |paddlepaddle |2020-12-15
cnnseg16 |3d_segmentation |lidar |paddlepaddle |2018-10-14
cnnseg128 |3d_segmentation |lidar |paddlepaddle |2020-06-17
cnnseg64 |3d_segmentation |lidar |paddlepaddle |2019-05-29
smoke |3d_detection |camera |paddlepaddle |2019-06-27
3d-yolo |3d_detection |camera |paddlepaddle |2019-12-08
denseline |lane_detection |camera |paddlepaddle |2019-05-29
darkSCNN |lane_detection |camera |paddlepaddle |2020-12-15
tl_detection |tl_detection |camera |paddlepaddle |2021-01-15
tl_recognition |tl_recognition |camera |paddlepaddle |2021-01-15
Then you can use the info command to show the details of the model.
$ amodel info point_pillars
name: point_pillars
date: 2020-12-15
task_type: 3d_detection
sensor_type: lidar
framework: paddlepaddle
model_files:
- name: pfe.onnx
size: 4125
- name: pts_backbone.zip
size: 16945051
- name: pts_bbox_head.zip
size: 121150
- name: pts_middle_encoder.zip
size: 3763
- name: pts_neck.zip
size: 2420625
- name: pts_voxel_encoder.zip
size: 17575
- name: rpn.onnx
size: 18300546
dataset:
- waymo
- kitti
- nusense
Also you can use the model's ID
amodel info 12 # 12 is point_pillars's ID
You can deploy the model through the install command.
# Install from local
amodel install xxx.zip
# Install from http
amodel install https://xxx.zip
You can delete the model installed in Apollo through the remove command.
amodel remove point_pillars
// or
amodel remove 12 # 12 is point_pillars's ID