def createScene(self, stepNum):
bundleFile = os.path.abspath(self.config.NVMFileName)
imgDir = os.path.abspath(self.config.SRCImageDirName)
outDir = os.path.abspath(self.config.NVMOutputDirName)
if False == os.path.isfile(bundleFile):
log("Step %d - SceneCreation cannot begin because the NVM file %s is missing." %
(stepNum, bundleFile))
log("Processing will not proceed.")
return -stepNum
if False == os.path.isdir(imgDir):
log("Step %d - SceneCreation cannot beging because the image directory %s cannot be found." %
(stepNum, imgDir))
log("Processing will not proceed.")
return -stepNum
log("Creating scene with bundleFile %s." % (bundleFile))
log("Output will be written to %s." % (outDir))
boxm2_batch.not_verbose()
boxm2_batch.register_processes()
boxm2_batch.register_datatypes()
# class used for python/c++ pointers in database
class dbvalue:
def __init__(self, index, type):
self.id = index # unsigned integer
self.type = type # string
# run process
boxm2_batch.init_process("boxm2BundleToSceneProcess")
boxm2_batch.set_input_string(0, bundleFile)
boxm2_batch.set_input_string(1, imgDir)
boxm2_batch.set_input_string(2, self.config.NVMAppModel)
boxm2_batch.set_input_string(3, self.config.NVMNobsModel)
boxm2_batch.set_input_int(4, 8)
boxm2_batch.set_input_string(5, outDir)
boxm2_batch.run_process()
(scene_id, scene_type) = boxm2_batch.commit_output(0)
uscene = dbvalue(scene_id, scene_type)
(scene_id, scene_type) = boxm2_batch.commit_output(1)
rscene = dbvalue(scene_id, scene_type)
return 0
# Recovering the correct offsets for rpc cameras
#
# s3_copy_files_from_local.py <local_folder> <name_ending> <hdfs_folder>
#
# all the files in <local_folder> as:
# <local_folder>/*<name_ending> will be copied over to <hdfs_working_dir>/<hdfs_folder> on hdfs
# e.g. python s3_copy_files_from_local.py ./crops _normalized.png world_dir
# <hdfs_working_dir> will be retrieved from <hdfs> manager
#
import boxm2_batch
boxm2_batch.register_processes()
boxm2_batch.register_datatypes()
import glob
import random
import math
import os
import shutil
import time
import sys
class dbvalue:
def __init__(self, index, type):
self.id = index # unsigned integer
self.type = type # string
print "create FS"
boxm2_batch.init_process("bhdfsCreateFSManagerProcess")
boxm2_batch.set_input_string(0, "default")
boxm2_batch.set_input_int(1, 0)
# Recovering the correct offsets for rpc cameras
#
# s4_prepare_stdin_txt.py <local_folder1> <name_ending1> <local folder2> <name_ending2> <hdfs file name> <hdfs_output_folder>
#
# prepare a file list with all the files in <local_folder1> as:
# <local_folder1>/*<name_ending1>
# a list such as
# <hdfs_working_dir>/<hdfs_output_folder>/<file name1> <hdfs_working_dir>/<hdfs_output_folder>/<file name2>
# and copy the file to <hdfs_working_dir>/<hdfs file name> ; <hdfs_working_dir> will be retrieved from <hdfs> manager
# e.g. python ./crops _normalized.png ./crops ./crops/exp_imgs _exp.png some_hdfs_folder/input_stdin.txt hadoop_output_folder
#
#
#
import boxm2_batch
boxm2_batch.register_processes()
boxm2_batch.register_datatypes()
import glob
import random
import math
import os
import shutil
import time
import sys
class dbvalue:
def __init__(self, index, type):
self.id = index # unsigned integer
self.type = type # string
def writeSceneFromBox(data_path, resolution, min_pt, max_pt, ntrees_x=64,ntrees_y=64,ntrees_z=64, max_num_lvls=4, appearance_model1 = "boxm2_mog3_grey", appearance_model2 = "boxm2_num_obs", appearance_model3 = "boxm2_sum_log_msg_pos", p_init=0.001,max_data_size=1500.0):
"""A function that takes the minimum and maximum points of a bounding box for the scene in real world coordinates
and partitions the space into the appropriate number of boxes given a user specified number of trees.
min_pt : A python list specifying the 3d position of the minimum corner of the box.
This is interpreted by boxm2 as the 3d origin of the scene.
max-pt : A python list specifying the 3d position of the maximum corner of the box.
ntrees : Number of trees in all dimensions (the python interface only supports symmetric trees thus far)
max_num_lvls : Maximum number of levels in the trees (a tree will have pow(2,max_num_lvls-1) possible cells)
appearance_model : A string indicating desired appearance model
occupancy_model : A string indicating desired occupancy model
max_data_size : Maximum Sizer of a block in megabytes. Determined by GPU memory. Recomment 650MB for 1GB card and 1.1GB for 1.5GM card
maximum_data_size : Maximum memory allowable for a superblock. The max_data_size is determined by GPU memory size.
For a 1 GB card, a maximum data size of 650MB is recommended.
For a 1.5GB card, a 1.1G maximum data size of 1.1GB is recommended.
p_init : Initial occupancy probability."""
boxm2_batch.register_processes();
boxm2_batch.register_datatypes();
tree_size=resolution*pow(2,max_num_lvls-1);
block_size_x=ntrees_x*tree_size;
block_size_y=ntrees_y*tree_size;
block_size_z=ntrees_z*tree_size;
xsize=max_pt[0]-min_pt[0];
ysize=max_pt[1]-min_pt[1];
zsize=max_pt[2]-min_pt[2];
print "zsize: %f" % zsize
print "block size: %f" % block_size_z
nblocks_x=int(round(xsize/block_size_x));
nblocks_y=int(round(ysize/block_size_y));
nblocks_z=int(round(zsize/block_size_z));
print "nblocks_z : %f" % nblocks_z
if(nblocks_x<=0):nblocks_x=1;
if(nblocks_y<=0):nblocks_y=1;
if(nblocks_z<=0):nblocks_z=1;
print '\t Number of blocks in the x dimension: %d' % nblocks_x
print '\t Number of blocks in the y dimension: %d' % nblocks_y
print '\t Number of blocks in the z dimension: %d' % nblocks_z
print("\t CREATING THE BOXM2_SCENE_SPTR")
boxm2_batch.init_process("boxm2CreateSceneProcess");
boxm2_batch.set_input_string(0,data_path);
boxm2_batch.set_input_string(1,appearance_model1);
boxm2_batch.set_input_string(2,appearance_model2);
boxm2_batch.set_input_string(3,appearance_model3);
boxm2_batch.set_input_float(4,min_pt[0]);
boxm2_batch.set_input_float(5,min_pt[1]);
boxm2_batch.set_input_float(6,min_pt[2]);
boxm2_batch.run_process();
(id,type)=boxm2_batch.commit_output(0);
scene=dbvalue(id,type);
for k in range(nblocks_z):
for j in range(nblocks_y):
for i in range(nblocks_x):
block_origin_x=min_pt[0]+i*block_size_x;
block_origin_y=min_pt[1]+j*block_size_y;
block_origin_z=min_pt[2]+k*block_size_z;
print '\t \t Creating block with id (%d,%d,%d) at origin (%s,%s,%s)' % (i,j,k,block_origin_x, block_origin_y, block_origin_z)
boxm2_batch.init_process("boxm2AddBlockProcess");
boxm2_batch.set_input_from_db(0,scene);
boxm2_batch.set_input_int(1,i);
boxm2_batch.set_input_int(2,j);
boxm2_batch.set_input_int(3,k);
boxm2_batch.set_input_unsigned(4,ntrees_x);
boxm2_batch.set_input_unsigned(5,ntrees_y);
boxm2_batch.set_input_unsigned(6,ntrees_z);
boxm2_batch.set_input_unsigned(7,max_num_lvls);
boxm2_batch.set_input_float(8,block_origin_x);
boxm2_batch.set_input_float(9,block_origin_y);
boxm2_batch.set_input_float(10,block_origin_z);
boxm2_batch.set_input_float(11,tree_size);
boxm2_batch.set_input_float(12,max_data_size);
boxm2_batch.set_input_float(13,p_init);
boxm2_batch.set_input_unsigned(14,1);
boxm2_batch.run_process();
boxm2_batch.init_process("boxm2WriteSceneXMLProcess")
boxm2_batch.set_input_from_db(0,scene);
boxm2_batch.set_input_string(1,"scene");
boxm2_batch.run_process();
def boxm2WriteSceneXML(data_path, resolution, origin, nblocks_x, nblocks_y,nblocks_z,ntrees=64,max_num_lvls=4, appearance_model = "boxm2_mog3_grey", occupancy_model = "boxm2_num_obs", max_data_size=650.0,p_init=0.001):
"""This is boxm2WriteSceneXML.py
Function to create a boxm2 scene XML file
Author: Brandon Mayer
Date: 4/6/2011
data_path : A directory to which the xml file should be saved
origin : A python list specifying the 3d origin (x,y,z)
nblocks_x : Number of superblocks in the x dimension
nblocks_y : Number of superblocks in the y dimension
nblocks_z : Number of superblocks in the z dimension
ntrees : Number of trees in all dimensions (the python interface only supports symmetric trees thus far)
max_num_lvls : Maximum number of levels in the trees (a tree will have pow(2,max_num_lvls-1) possible cells)
appearance_model : A string indicating desired appearance model
occupancy_model : A string indicating desired occupancy model
maximum_data_size : Maximum memory allowable for a superblock. The max_data_size is determined by GPU memory size.
For a 1 GB card, a maximum data size of 650MB is recommended.
For a 1.5GB card, a 1.1G maximum data size of 1.1GB is recommended.
p_init : Initial occupancy probability."""
boxm2_batch.register_processes();
boxm2_batch.register_datatypes();
tree_size=resolution*pow(2,max_num_lvls-1);
block_size=ntrees*tree_size;
print("\t Creating the boxm2_scene_sptr")
boxm2_batch.init_process("boxm2CreateSceneProcess");
boxm2_batch.set_input_string(0,data_path);
boxm2_batch.set_input_string(1,appearance_model);
boxm2_batch.set_input_string(2,occupancy_model);
boxm2_batch.set_input_float(3,origin[0]);
boxm2_batch.set_input_float(4,origin[1]);
boxm2_batch.set_input_float(5,origin[2]);
boxm2_batch.run_process();
(id,type) = boxm2_batch.commit_output(0);
scene = dbvalue(id,type);
for i in range(nblocks_x):
for j in range(nblocks_y):
for k in range(nblocks_z):
block_origin_x=origin[0]+i*block_size;
block_origin_y=origin[1]+j*block_size;
block_origin_z=origin[2]+k*block_size;
print '\t Creating block with id (%d,%d,%d) at origin (%s,%s,%s)' % (i,j,k,block_origin_x, block_origin_y, block_origin_z)
boxm2_batch.init_process("boxm2AddBlockProcess");
boxm2_batch.set_input_from_db(0,scene);
boxm2_batch.set_input_unsigned(1,i);
boxm2_batch.set_input_unsigned(2,j);
boxm2_batch.set_input_unsigned(3,k);
boxm2_batch.set_input_unsigned(4,ntrees);
boxm2_batch.set_input_unsigned(5,ntrees);
boxm2_batch.set_input_unsigned(6,ntrees);
boxm2_batch.set_input_unsigned(7,max_num_lvls);
boxm2_batch.set_input_float(8,block_origin_x);
boxm2_batch.set_input_float(9,block_origin_y);
boxm2_batch.set_input_float(10,block_origin_z);
boxm2_batch.set_input_float(11,tree_size);
boxm2_batch.set_input_float(12,max_data_size);
boxm2_batch.set_input_float(13,p_init);
boxm2_batch.run_process();
boxm2_batch.init_process("boxm2WriteSceneXMLProcess")
boxm2_batch.set_input_from_db(0,scene)
boxm2_batch.run_process();
def writeSceneFromBox(data_path,
resolution,
min_pt,
max_pt,
ntrees_x=64,
ntrees_y=64,
ntrees_z=64,
max_num_lvls=4,
appearance_model1="boxm2_mog3_grey",
appearance_model2="boxm2_num_obs",
appearance_model3="boxm2_sum_log_msg_pos",
p_init=0.001,
max_data_size=1500.0):
"""A function that takes the minimum and maximum points of a bounding box for the scene in real world coordinates
and partitions the space into the appropriate number of boxes given a user specified number of trees.
min_pt : A python list specifying the 3d position of the minimum corner of the box.
This is interpreted by boxm2 as the 3d origin of the scene.
max-pt : A python list specifying the 3d position of the maximum corner of the box.
ntrees : Number of trees in all dimensions (the python interface only supports symmetric trees thus far)
max_num_lvls : Maximum number of levels in the trees (a tree will have pow(2,max_num_lvls-1) possible cells)
appearance_model : A string indicating desired appearance model
occupancy_model : A string indicating desired occupancy model
max_data_size : Maximum Sizer of a block in megabytes. Determined by GPU memory. Recomment 650MB for 1GB card and 1.1GB for 1.5GM card
maximum_data_size : Maximum memory allowable for a superblock. The max_data_size is determined by GPU memory size.
For a 1 GB card, a maximum data size of 650MB is recommended.
For a 1.5GB card, a 1.1G maximum data size of 1.1GB is recommended.
p_init : Initial occupancy probability."""
boxm2_batch.register_processes()
boxm2_batch.register_datatypes()
tree_size = resolution * pow(2, max_num_lvls - 1)
block_size_x = ntrees_x * tree_size
block_size_y = ntrees_y * tree_size
block_size_z = ntrees_z * tree_size
xsize = max_pt[0] - min_pt[0]
ysize = max_pt[1] - min_pt[1]
zsize = max_pt[2] - min_pt[2]
print "zsize: %f" % zsize
print "block size: %f" % block_size_z
nblocks_x = int(round(xsize / block_size_x))
nblocks_y = int(round(ysize / block_size_y))
nblocks_z = int(round(zsize / block_size_z))
print "nblocks_z : %f" % nblocks_z
if (nblocks_x <= 0): nblocks_x = 1
if (nblocks_y <= 0): nblocks_y = 1
if (nblocks_z <= 0): nblocks_z = 1
print '\t Number of blocks in the x dimension: %d' % nblocks_x
print '\t Number of blocks in the y dimension: %d' % nblocks_y
print '\t Number of blocks in the z dimension: %d' % nblocks_z
print("\t CREATING THE BOXM2_SCENE_SPTR")
boxm2_batch.init_process("boxm2CreateSceneProcess")
boxm2_batch.set_input_string(0, data_path)
boxm2_batch.set_input_string(1, appearance_model1)
boxm2_batch.set_input_string(2, appearance_model2)
boxm2_batch.set_input_string(3, appearance_model3)
boxm2_batch.set_input_float(4, min_pt[0])
boxm2_batch.set_input_float(5, min_pt[1])
boxm2_batch.set_input_float(6, min_pt[2])
boxm2_batch.run_process()
(id, type) = boxm2_batch.commit_output(0)
scene = dbvalue(id, type)
for k in range(nblocks_z):
for j in range(nblocks_y):
for i in range(nblocks_x):
block_origin_x = min_pt[0] + i * block_size_x
block_origin_y = min_pt[1] + j * block_size_y
block_origin_z = min_pt[2] + k * block_size_z
print '\t \t Creating block with id (%d,%d,%d) at origin (%s,%s,%s)' % (
i, j, k, block_origin_x, block_origin_y, block_origin_z)
boxm2_batch.init_process("boxm2AddBlockProcess")
boxm2_batch.set_input_from_db(0, scene)
boxm2_batch.set_input_int(1, i)
boxm2_batch.set_input_int(2, j)
boxm2_batch.set_input_int(3, k)
boxm2_batch.set_input_unsigned(4, ntrees_x)
boxm2_batch.set_input_unsigned(5, ntrees_y)
boxm2_batch.set_input_unsigned(6, ntrees_z)
boxm2_batch.set_input_unsigned(7, max_num_lvls)
boxm2_batch.set_input_float(8, block_origin_x)
boxm2_batch.set_input_float(9, block_origin_y)
boxm2_batch.set_input_float(10, block_origin_z)
boxm2_batch.set_input_float(11, tree_size)
boxm2_batch.set_input_float(12, max_data_size)
boxm2_batch.set_input_float(13, p_init)
boxm2_batch.set_input_unsigned(14, 1)
boxm2_batch.run_process()
boxm2_batch.init_process("boxm2WriteSceneXMLProcess")
boxm2_batch.set_input_from_db(0, scene)
boxm2_batch.set_input_string(1, "scene")
boxm2_batch.run_process()
