# needed packages
require(e1071)
require(randomForest)
require(automap)
require(biOps)
require(DMwR)
require(compiler)
require(parallel)

# load our scripts files, where the real functions are implemented.
source("spatialMatrixRegression.R")
source("lista.R")
source("spatialModels.R")
source("funcoes.R")


# transform the picture to an R matrix...
nina_face <- readJpeg("data/nina_face.jpg")
nina_face <- unclass(nina_face)
save(nina_face, file="data/nina_face.RData")

# randomly create a regression matrix according with the paper.
# percents: represents the percentage of missing data, randomly selected
# qtd.samples: how many times we repeat this process
# could take a while... In multicore machines increase the qtd.cores parameter!
obj_nina_face <- createObj(mat=nina_face, percents=.6, qtd.samples=1, radius1=10, radius2=20, radius3=30, minElementR1=2, minElementR2=5, minElementR3=10, qtd.cores=1)
save(obj_nina_face, file="data/obj_nina_face.RData")
# save the picture, replacing the missing locations (NA) to white pixel (255).
saveNAJpg(obj_nina_face)

# learn the models 
DEBUG <- T
train.trees(obj_nina_face, "obj_nina_face")
train.rfs(obj_nina_face, "obj_nina_face")
train.svms(obj_nina_face, "obj_nina_face")
train.base.models(obj_nina_face, "obj_nina_face")
learn.autokrige.saving(obj_nina_face, "obj_nina_face", nmax=90)

# transforming the predictions to jpg
load("data/obj_nina_face.RData")
load("data/obj_nina_face_rf.RData")
saveModelJpg(obj_nina_face, rf, "rf", file.name="nina_face_rf")

load("data/obj_nina_face_svm.RData")
saveModelJpg(obj_nina_face, svm, "svm", file.name="nina_face_svm")

load("data/obj_nina_face_tree.RData")
saveModelJpg(obj_nina_face, tree, "tree", file.name="nina_face_tree")

load("data/obj_nina_face_base.RData")
saveModelJpg(obj_nina_face, obj_nina_face.m1, "m1", type="base", file.name="nina_face_m1")
saveModelJpg(obj_nina_face, obj_nina_face.m2, "m2", type="base", file.name="nina_face_m2")
saveModelJpg(obj_nina_face, obj_nina_face.m3, "m3", type="base", file.name="nina_face_m3")

saveModelJpg(obj_nina_face, obj_nina_face.w1, "w1", type="base", file.name="nina_face_w1")
saveModelJpg(obj_nina_face, obj_nina_face.w2, "w2", type="base", file.name="nina_face_w2")
saveModelJpg(obj_nina_face, obj_nina_face.w3, "w3", type="base", file.name="nina_face_w3")

load("data/obj_nina_face_autoKrige.RData")
saveModelJpg(obj_nina_face, model, "kc", file.name="nina_face_kc")

# Coliseum

# transform the picture to R matrix...
coliseum <- readJpeg("data/coliseum.jpg")
coliseum <- unclass(coliseum)
save(coliseum, file="data/coliseum.RData")

# randomly create an regression matrix according with the paper.
# percents: represents the percentage of missing data, randomly selected
# qtd.samples: how many times we repeat this process
# could take a while... In multicore machines increase the qtd.cores parameter!
obj_coliseum <- createObj(mat=coliseum, percents=.6, qtd.samples=1, radius1=10, radius2=20, radius3=30, minElementR1=2, minElementR2=5, minElementR3=10, qtd.cores=1)
save(obj_coliseum, file="data/obj_coliseum.RData")
# save the picture, replacing the missing locations (NA) to white pixel (255).
saveNAJpg(obj_coliseum, file.name="coliseum")

# learn the models 
DEBUG <- T
train.trees(obj_coliseum, "obj_coliseum")
train.rfs(obj_coliseum, "obj_coliseum")
train.svms(obj_coliseum, "obj_coliseum")
train.base.models(obj_coliseum, "obj_coliseum")
learn.autokrige.saving(obj_coliseum, "obj_coliseum", nmax=90)

# transforming the predictions to jpg
load("data/obj_coliseum.RData")
load("data/obj_coliseum_rf.RData")
saveModelJpg(obj_coliseum, rf, "rf", file.name="coliseum_rf")

load("data/obj_coliseum_svm.RData")
saveModelJpg(obj_coliseum, svm, "svm", file.name="coliseum_svm")

load("data/obj_coliseum_tree.RData")
saveModelJpg(obj_coliseum, tree, "tree", file.name="coliseum_tree")

load("data/obj_coliseum_base.RData")
saveModelJpg(obj_coliseum, obj_coliseum.m1, "m1", type="base", file.name="coliseum_m1")
saveModelJpg(obj_coliseum, obj_coliseum.m2, "m2", type="base", file.name="coliseum_m2")
saveModelJpg(obj_coliseum, obj_coliseum.m3, "m3", type="base", file.name="coliseum_m3")

saveModelJpg(obj_coliseum, obj_coliseum.w1, "w1", type="base", file.name="coliseum_w1")
saveModelJpg(obj_coliseum, obj_coliseum.w2, "w2", type="base", file.name="coliseum_w2")
saveModelJpg(obj_coliseum, obj_coliseum.w3, "w3", type="base", file.name="coliseum_w3")

load("data/obj_coliseum_autoKrige.RData")
saveModelJpg(obj_coliseum, model, "kc", file.name="coliseum_kc")