How to Install Caffe for SSD vision detector

Updated as of 2018-07-04

1. Complete the pre-requisites.

2. Clone the SSD Caffe fork in your home directory (CMake files will be looking for it there).

% git clone -b ssd https://github.com/weiliu89/caffe.git ssdcaffe
% cd ssdcaffe
% git checkout 4817bf8b4200b35ada8ed0dc378dceaf38c539e4

1. Follow the authors’ instructions to complete the pre-requisites for compilation.

2. Compile Caffe:

make && make distribute

1. Download pre-trained models as provided at https://github.com/weiliu89/caffe/tree/ssd, or use your own.

Once compiled, run from the terminal, or launch from RunTimeManager:

% roslaunch vision_ssd_detect vision_ssd_detect network_definition_file:=/PATH/TO/deploy.prototxt pretrained_model_file:=/PATH/TO/model.caffemodel

Remember to modify the launch file located inside AUTOWARE_BASEDIR/ros/src/computing/perception/detection/vision_detector/packages/vision_ssd_detect/launch/vision_ssd_detect.launch and point the network and pre-trained models to your paths.

Launch file params

Notes

Remember to modify the deploy.prototxt to use the FULL PATH to the VOC label prototxt file.

Open the file $SSDCAFFEPATH/models/VGGNet/VOC0712/SSD_512x512/deploy.prototxt

Change line 1803 to point to the full path of the labelmap_voc.prototxt file.

Changelog for package vision_ssd_detect

1.11.0 (2019-03-21)

• [fix] Install commands for all the packages (#1861)

  • Initial fixes to detection, sensing, semantics and utils

  • fixing wrong filename on install command

  • Fixes to install commands

  • Hokuyo fix name

  • Fix obj db

  • Obj db include fixes

  • End of final cleaning sweep

  • Incorrect command order in runtime manager

  • Param tempfile not required by runtime_manager

  • • Fixes to runtime manager install commands

  • Remove devel directory from catkin, if any

  • Updated launch files for robosense

  • Updated robosense

  • Fix/add missing install (#1977)

  • Added launch install to lidar_kf_contour_track

  • Added install to op_global_planner

  • Added install to way_planner

  • Added install to op_local_planner

  • Added install to op_simulation_package

  • Added install to op_utilities

  • Added install to sync

  • • Improved installation script for pointgrey packages

  • Fixed nodelet error for gmsl cameras

  • USe install space in catkin as well

  • add install to catkin

  • Fix install directives (#1990)

  • Fixed installation path

  • Fixed params installation path

  • Fixed cfg installation path

  • Delete cache on colcon_release
• Fix license notice in corresponding package.xml
• Fix/remove obsolete range fusion (#1884)
  • Remove obsolete range_fusion
  • Removed range_fusion ui entry
  • Remove obsolete lidar_tracker_libs_fusion
  • Removed fusion dependency from vision_ssd_detect package
• Initial release of object filter
• Contributors: Abraham Monrroy, Abraham Monrroy Cano, Servando, amc-nu

1.10.0 (2019-01-17)

• Fixes for catkin_make
• [fix] SSD detector, cmake colcon (#1837)

  • Fixes for new colcon script on ssd cuda based node

  • Fixed to RTM and darknet launch files

  • catkin_fix

  • • catkin & colcon build successfully
  • reverted back run to devel space (for the time being)
• Switch to Apache 2 license (develop branch) (#1741)
  • Switch to Apache 2

  * Replace BSD-3 license header with Apache 2 and reassign copyright to the Autoware Foundation.

  • Update license on Python files
  • Update copyright years

File truncated at 100 lines see the full file