Several source files were created in this package. Although most of them are very similar, each is to be used in a specific situation. Three binaries are created when the code is compiled. A server, a client and a trainer. Note that the CMakeLists.txt file has to be edited according to the source files you want to use.
The server files are used to generate the bins that send images over to a ROS Topic.
serverimg_singleimg.cpp<b> This source file selects an image from file and sends that same image repeatedly over to a ROS Topic while it is running. It is useful for developing and test situations.
To use it, edit the source code to add path to the image you want to work with and make sure the CMakeLists.txt file is set to use this source file to produce the binary.
serverimg_folder.cpp<b> The most useful server source code. This code goes through the images contained in a folder and sends them over to the ROS Topic.
It is useful to work with large datasets.
The client subscribes to the ROS Topic to which the server is publishing. A received image triggers a callback function that processes the image. Make sure that the ROS Topic names are the same in the server and in the client, or else the program will ot run.
clientimg_channels.cpp<b>
This client was the product of the first stages of the developed work and it only computes the channels. Also, the channels are computed in the main function, rather than in a different function which is the case of the other clients. Easy access to the channels facilitated the development in the first stages of the work.
clientimg_deprecated.cpp<b>
This source file is deprecated and has no utility.
clientimg_train_pos.cpp<b>
This client is used to write the features of positive windows on a .csv file, which is later used for training. Edit this file to the desired filename on which the program will write. Be sure that the server is set to go through a folder of positive windows. No detection is performed yet.
clientimg_train_neg.cpp<b>
This client goes through a full image and selects 5 random windows to write its features on a .csv file, which is later used for training. Edit this file to set the desired filename on which the program will write. Be sure that the server is set to go through a folder of negative images (images where no pedestrians appear).
clientimg_detect.cpp<b>
This is the final client that is used to perform detection. It outputs an image with green rectangles around detected objects. This file is also used when the developer needs to perform bootstrapping, however, this operation is explained in another section.
boost_error.cpp<b>
Not useful anymore.