diff --git a/compute_features.py b/compute_features.py
index cee73a35acd1f93f49d67e949c93eb27ee43fbc4..958e47f89dca4ea4ebc04d910b9204f8bcd2dca0 100644
--- a/compute_features.py
+++ b/compute_features.py
@@ -21,7 +21,7 @@ def process_compute(data_name, camera, camera_resized, new_size, Nb_camera, Worl
destination_folders_list = {}
for scenario_loop in scenarios:
- destination_folders_list[scenario_loop] = [f"Generated_{scenario_loop}_Testing", f"Generated_{scenario_loop}_Evaluating", f"Generated_{scenario_loop}_Training" ]
+ destination_folders_list[scenario_loop] = [f"Generated_{scenario_loop}_Testing", f"Generated_{scenario_loop}_Evaluating", f"Generated_{scenario_loop}_Training", f"Generated_{scenario_loop}_dont_save" ]
list_count_categories = {}
for scenario_loop in scenarios :
@@ -33,10 +33,9 @@ def process_compute(data_name, camera, camera_resized, new_size, Nb_camera, Worl
for i in range(World_begin, World_begin + Nb_world): # worlds
- print(i)
- if i > 4 :
+ if i > 8000 :
destination_folders["Worlds"] = f"Generated_Worlds_Testing"
- elif i > 3 :
+ elif i > 6000 :
destination_folders["Worlds"] = f"Generated_Worlds_Evaluating"
else :
destination_folders["Worlds"] = f"Generated_Worlds_Training"
@@ -53,14 +52,14 @@ def process_compute(data_name, camera, camera_resized, new_size, Nb_camera, Worl
else :
destination_folders["Cameras"] = f"Generated_Cameras_Training"
- if i > 4 and j > 12 :
+ if i > 8000 and j > 12 :
destination_folders["Mix_all"] = f"Generated_Mix_all_Testing"
- elif i > 3 and i <=4 and j > 9 and j >= 12 :
+ elif i > 6000 and i <=8000 and j > 9 and j >= 12 :
destination_folders["Mix_all"] = f"Generated_Mix_all_Evaluating"
- elif i <= 3 and j <= 9 :
+ elif i <= 6000 and j <= 9 :
destination_folders["Mix_all"] = f"Generated_Mix_all_Training"
else :
- destination_folders["Mix_all"] = "dont_save"
+ destination_folders["Mix_all"] = f"Generated_Mix_all_dont_save"
categories_occ_array, categories_array = compute_id_good_occ(data_name, p, categories_instance_array_id_to_cat, categories_instance_array_cat_to_id, occ_target)
@@ -79,10 +78,10 @@ def process_compute(data_name, camera, camera_resized, new_size, Nb_camera, Worl
raise TypeError("size of datas are differents !!")
for scenario_loop in scenarios:
- if not destination_folders[scenario_loop] == "dont_save" :
- if os.path.isfile(f'{data_name}/{destination_folders[scenario_loop]}/Count_{p-1}.json'):
- with open(f'{data_name}/{destination_folders[scenario_loop]}/Count_{p-1}.json') as f:
- list_count_categories[scenario_loop][destination_folders[scenario_loop]] = json.load(f)
+ for destination_folder_loop in destination_folders_list[scenario_loop] :
+ if os.path.isfile(f'{data_name}/{destination_folder_loop}/Count_{p-1}.json'):
+ with open(f'{data_name}/{destination_folder_loop}/Count_{p-1}.json') as f:
+ list_count_categories[scenario_loop][destination_folder_loop] = json.load(f)
#res_all = []
for categories in list_categories:
@@ -112,17 +111,17 @@ def process_compute(data_name, camera, camera_resized, new_size, Nb_camera, Worl
meta['camera'] = f"grabber_{j}"
meta['occlusion'] = occ_target
meta['Nb_instance_category'] = Nb_instance
-
- if not os.path.isfile(f'{data_name}/{destination_folders[scenario_loop]}/{categories}/Meta_Gen/{categories}.json'):
- with open(f'{data_name}/{destination_folders[scenario_loop]}/{categories}/Meta_Gen/{categories}.json', mode='w') as f:
+
+ if not os.path.isfile(f'{data_name}/{destination_folders[scenario_loop]}/{categories}/Meta_Gen/{categories}_{p}.json'):
+ with open(f'{data_name}/{destination_folders[scenario_loop]}/{categories}/Meta_Gen/{categories}_{p}.json', mode='w') as f:
feeds = {}
feeds[meta['id_generated']]=meta
f.write(json.dumps(feeds, indent=2))
else:
- with open(f'{data_name}/{destination_folders[scenario_loop]}/{categories}/Meta_Gen/{categories}.json') as feedsjson:
+ with open(f'{data_name}/{destination_folders[scenario_loop]}/{categories}/Meta_Gen/{categories}_{p}.json') as feedsjson:
feeds = json.load(feedsjson)
feeds[meta['id_generated']]=meta
- with open(f'{data_name}/{destination_folders[scenario_loop]}/{categories}/Meta_Gen/{categories}.json', mode='w') as f:
+ with open(f'{data_name}/{destination_folders[scenario_loop]}/{categories}/Meta_Gen/{categories}_{p}.json', mode='w') as f:
f.write(json.dumps(feeds, indent=4))
if (Nb_instance_occ == 1 and Nb_instance == 0 ): # condition of only one instance of occ >= 0.5 and no other < 0.05
@@ -141,24 +140,18 @@ def process_compute(data_name, camera, camera_resized, new_size, Nb_camera, Worl
for scenario_loop in scenarios:
if not destination_folders[scenario_loop] == "dont_save" :
np.save(f'{data_name}/{destination_folders[scenario_loop]}/{categories}/Pose_transformed/{p}.npy', num_arr) # save
- else:
- continue
if data_Bbox_2d[k]['id'] == categories_occ_array[categories][0]:
bbox = bbox_2d(data_Bbox_2d[k])
for scenario_loop in scenarios:
if not destination_folders[scenario_loop] == "dont_save" :
np.savetxt(f'{data_name}/{destination_folders[scenario_loop]}/{categories}/Bbox/{p}.txt', np.array(bbox).reshape((1, 4))) # save
- else:
- continue
if data_Bbox_3d[k]['id'] == categories_occ_array[categories][0]:
bbox3d_size = data_Bbox_3d[k]['bbox']['size']
for scenario_loop in scenarios:
if not destination_folders[scenario_loop] == "dont_save" :
np.savetxt(f'{data_name}/{destination_folders[scenario_loop]}/{categories}/Bbox_3d_Gen/{p}.txt', bbox3d_size) # save
- else:
- continue
id = categories_occ_array[categories][0]
img = cv2.imread(f"{data_name}/Instance_Segmentation/{p}.png", cv2.IMREAD_UNCHANGED) # plt.imread(path)
@@ -174,6 +167,8 @@ def process_compute(data_name, camera, camera_resized, new_size, Nb_camera, Worl
image = cv2.imread(f"{data_name}/{destination_folders[scenario_loop]}/{categories}/Instance_Mask/{p}.png", 0)
image = image/255.0
contours, _ = cv2.findContours(image.astype(np.uint8), cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
+ if len(contours) > 1 :
+ continue
for l in range(len(contours[0])):
x = contours[0][l][0][0]/640.0
res.append(x)
@@ -246,5 +241,6 @@ def process_compute(data_name, camera, camera_resized, new_size, Nb_camera, Worl
f.write(json.dumps(list_count_categories[scenario_loop][destination_folder_loop], indent=4))
with open(f'{data_name}/Count_{p}.json', mode='w') as f:
f.write(json.dumps(list_count_categories, indent=4))
- #print(list_count_categories)
+ print(list_count_categories)
+ print(f'{data_name}/{destination_folder_loop}/Count_{p}.json')
diff --git a/main.py b/main.py
index a746677cadde9540c6372750d97009852a548c6c..8405c9c241a7f9c43bcce957654d0e6a3c6e221c 100644
--- a/main.py
+++ b/main.py
@@ -36,9 +36,9 @@ def generate_folders(name, list_categories, scenario):
is_exist2 = os.path.exists(f"{name}/Generated_{scenario}_Testing/{cat}/{f}")
if not is_exist2:
os.makedirs(f"{name}/Generated_{scenario}_Testing/{cat}/{f}")
- is_exist2 = os.path.exists(f"{name}/dont_save/{cat}/{f}")
+ is_exist2 = os.path.exists(f"{name}/Generated_{scenario}_dont_save/{cat}/{f}")
if not is_exist2:
- os.makedirs(f"{name}/dont_save/{cat}/{f}")
+ os.makedirs(f"{name}/Generated_{scenario}_dont_save/{cat}/{f}")
@@ -53,13 +53,15 @@ def calc_pts_diameter2(pts):
diameter = np.max(dists)
return diameter
-if __name__ == '__main__':
-
+if __name__ == '__main__':
# Create the parser
parser = argparse.ArgumentParser()
# Add an argument
parser.add_argument('--Nb_worlds', type=int, required=True)
parser.add_argument('--World_begin', type=int, required=True)
+ parser.add_argument('--dataset_id', type=str, required=True)
+ parser.add_argument('--rearrange', type=bool, required=True)
+ parser.add_argument('--compute', type=bool, required=True)
# Parse the argument
args = parser.parse_args()
@@ -69,14 +71,20 @@ if __name__ == '__main__':
Categories = [] # to read
Nb_instance = 1
occ_target = 0.5
- dataset_src = "/media/gduret/DATA/dataset/s2rg/Fruits_all_medium/data"
+
+ dataset_src = f"/gpfsscratch/rech/uli/ubn15wo/data{args.dataset_id}"
+ #dataset_src = "/media/mahmoud/E/Fruits_easy/data"
+ #dataset_src = "/media/gduret/DATA/dataset/s2rg/Fruits_all_medium/data"
+
choice = "low" # depth of rgb resolution datas
data_options = {"high": "ground_truth_rgb",
"low": "ground_truth_depth"}
dataset_type = data_options[choice]
- dataset_name = f"GUIMOD_{choice}"
- list_categories = ["banana1", "kiwi1", "pear2", "strawberry1", "apricot", "orange2", "peach1", "lemon2", "apple2" ]
+ dataset_name = f"/gpfsscratch/rech/uli/ubn15wo/GUIMOD_New_{choice}_{args.dataset_id}"
+ #dataset_name = f"/gpfsscratch/rech/uli/ubn15wo/dataset_new{args.dataset_id}/s2rg/Fruits_all_medium/GUIMOD_{choice}"
+ list_categories = ["banana1", "kiwi1", "pear2", "apricot", "orange2", "peach1", "lemon2", "apple2" ]
Nb_camera = 15
+ #Nb_world = 10000
generate_folders(dataset_name, list_categories, scenario)
@@ -103,19 +111,20 @@ if __name__ == '__main__':
#np.savetxt(f'{dataset_name}/Generated/camera_{choice}.txt', camera)
- reform_data(dataset_src, dataset_name, dataset_type, Nb_camera, args.World_begin, args.Nb_worlds)
+ if args.rearrange :
+ reform_data(dataset_src, dataset_name, dataset_type, Nb_camera, args.World_begin, args.Nb_worlds)
- list_categories = ["banana1", "kiwi1", "pear2", "strawberry1", "apricot", "orange2", "peach1", "lemon2", "apple2" ]
objs = {"banana1": [ 0.02949700132012367249, 0.1511049866676330566, 0.06059300713241100311 ],
"kiwi1": [ 0.04908600077033042908, 0.07206099480390548706, 0.04909799993038177490 ],
"pear2": [ 0.06601099669933319092, 0.1287339925765991211, 0.06739201396703720093 ],
- "strawberry1": [0.01698100194334983826, 0.02203200198709964752, 0.01685700193047523499],
"apricot": [0.04213499650359153748, 0.05482299625873565674, 0.04333199933171272278],
"orange2": [ 0.07349500805139541626, 0.07585700601339340210, 0.07458199560642242432 ],
"peach1": [ 0.07397901266813278198, 0.07111301273107528687, 0.07657301425933837891 ],
"lemon2": [0.04686100035905838013, 0.04684200137853622437, 0.07244800776243209839],
"apple2": [0.05203099921345710754, 0.04766000062227249146, 0.05089000239968299866]}
+ # "strawberry1": [0.01698100194334983826, 0.02203200198709964752, 0.01685700193047523499],
+
for categories in list_categories:
point_cloud = f"Models/{categories}/{categories.lower()}.ply"
pcd = o3d.io.read_point_cloud(point_cloud)
@@ -133,5 +142,6 @@ if __name__ == '__main__':
bbox = get_3D_bbox(ext)
np.savetxt(f'{dataset_name}/Generated/{categories}/{categories}_bbox_3d.txt', bbox) # save
- process_compute(dataset_name, camera, new_camera, new_size, Nb_camera, args.World_begin, args.Nb_worlds, list_categories, occ_target, False)
+ if args.compute :
+ process_compute(dataset_name, camera, new_camera, new_size, Nb_camera, args.World_begin, args.Nb_worlds, list_categories, occ_target, False)
diff --git a/main_rearrange.py b/main_rearrange.py
new file mode 100644
index 0000000000000000000000000000000000000000..65a10dfa10f52e5101a426eca001aeea07311039
--- /dev/null
+++ b/main_rearrange.py
@@ -0,0 +1,142 @@
+import os
+import numpy as np
+from prepare_data import reform_data
+from fps_alg import apply_fps
+from bbox_3d import get_3D_bbox
+from compute_features import process_compute
+import open3d as o3d
+from scipy.spatial import distance
+import argparse
+
+def generate_folders(name, list_categories, scenario):
+ is_exist = os.path.exists(name)
+ if not is_exist:
+ os.mkdir(name)
+ folders = ["RGB", "RGB_Gen", "RGB_resized", "Meta_Gen", "Depth", "Mask", "Meta", "Pose", "Bbox_2d", "Bbox_2d_loose", "Bbox_3d", "Bbox_3d_Gen", "Instance_Segmentation", "Semantic_Segmentation", "Instance_Mask", "Instance_Mask_resized", "Occlusion", "Models", "Pose_transformed", "Bbox", "FPS", "FPS_resized"]
+ for f in folders:
+ is_exist = os.path.exists(f"{name}/{f}")
+ if not is_exist:
+ if f not in ["RGB_Gen", "RGB_resized", "Instance_Mask", "Instance_Mask_resized", "Meta_Gen", "Models", "Pose_transformed", "Bbox", "Bbox_3d_Gen", "FPS" , "FPS_resized"]:
+ os.mkdir(f"{name}/{f}")
+ else:
+ for cat in list_categories:
+ is_exist2 = os.path.exists(f"{name}/Generated/{cat}")
+ if not is_exist2:
+ os.makedirs(f"{name}/Generated/{cat}")
+ is_exist2 = os.path.exists(f"{name}/Generated/{cat}/Pose_transformed")
+ if not is_exist2:
+ os.makedirs(f"{name}/Generated/{cat}/Pose_transformed")
+ for scenario in ["Worlds", "Cameras", "Mix_all"] :
+ is_exist2 = os.path.exists(f"{name}/Generated_{scenario}_Training/{cat}/{f}")
+ if not is_exist2:
+ os.makedirs(f"{name}/Generated_{scenario}_Training/{cat}/{f}")
+ is_exist2 = os.path.exists(f"{name}/Generated_{scenario}_Evaluating/{cat}/{f}")
+ if not is_exist2:
+ os.makedirs(f"{name}/Generated_{scenario}_Evaluating/{cat}/{f}")
+ is_exist2 = os.path.exists(f"{name}/Generated_{scenario}_Testing/{cat}/{f}")
+ if not is_exist2:
+ os.makedirs(f"{name}/Generated_{scenario}_Testing/{cat}/{f}")
+ is_exist2 = os.path.exists(f"{name}/dont_save/{cat}/{f}")
+ if not is_exist2:
+ os.makedirs(f"{name}/dont_save/{cat}/{f}")
+
+
+
+def calc_pts_diameter2(pts):
+ """Calculates the diameter of a set of 3D points (i.e. the maximum distance
+ between any two points in the set). Faster but requires more memory than
+ calc_pts_diameter.
+ :param pts: nx3 ndarray with 3D points.
+ :return: The calculated diameter.
+ """
+ dists = distance.cdist(pts, pts, 'euclidean')
+ diameter = np.max(dists)
+ return diameter
+
+if __name__ == '__main__':
+ # Create the parser
+ parser = argparse.ArgumentParser()
+ # Add an argument
+ parser.add_argument('--Nb_worlds', type=int, required=True)
+ parser.add_argument('--World_begin', type=int, required=True)
+ parser.add_argument('--dataset_id', type=str, required=True)
+ # Parse the argument
+ args = parser.parse_args()
+
+ scenario = "Worlds"
+
+ ### parameters ###
+ Categories = [] # to read
+ Nb_instance = 1
+ occ_target = 0.5
+
+ dataset_src = f"/gpfsscratch/rech/uli/ubn15wo/data{args.dataset_id}"
+ #dataset_src = "/media/mahmoud/E/Fruits_easy/data"
+ #dataset_src = "/media/gduret/DATA/dataset/s2rg/Fruits_all_medium/data"
+
+ choice = "low" # depth of rgb resolution datas
+ data_options = {"high": "ground_truth_rgb",
+ "low": "ground_truth_depth"}
+ dataset_type = data_options[choice]
+ dataset_name = f"/gpfsscratch/rech/uli/ubn15wo/dataset{args.dataset_id}/s2rg/Fruits_all_medium/GUIMOD_{choice}"
+ list_categories = ["banana1", "kiwi1", "pear2", "strawberry1", "apricot", "orange2", "peach1", "lemon2", "apple2" ]
+ Nb_camera = 15
+ #Nb_world = 10000
+
+ generate_folders(dataset_name, list_categories, scenario)
+
+ if choice == 'high':
+ camera = np.matrix([[1386.4138492513919, 0.0, 960.5],
+ [0.0, 1386.4138492513919, 540.5],
+ [0.0, 0.0, 1.0]])
+ # (640/1920 = 1 / 3), (480/1080 = 4 / 9)
+ trans = np.matrix([[1 / 3, 0.0, 0.0],
+ [0.0, (4 / 9), 0.0],
+ [0.0, 0.0, 1.0]])
+ elif choice == 'low':
+ camera = np.matrix([[1086.5054444841007, 0.0, 640.5],
+ [0.0, 1086.5054444841007, 360.5],
+ [0.0, 0.0, 1.0]])
+ #
+ trans = np.matrix([[0.5, 0.0, 0.0],
+ [0.0, (2 / 3), 0.0],
+ [0.0, 0.0, 1.0]])
+
+ new_size = (640, 480)
+
+ new_camera = trans @ camera
+
+ #np.savetxt(f'{dataset_name}/Generated/camera_{choice}.txt', camera)
+
+ reform_data(dataset_src, dataset_name, dataset_type, Nb_camera, args.World_begin, args.Nb_worlds)
+
+ list_categories = ["banana1", "kiwi1", "pear2", "strawberry1", "apricot", "orange2", "peach1", "lemon2", "apple2" ]
+ objs = {"banana1": [ 0.02949700132012367249, 0.1511049866676330566, 0.06059300713241100311 ],
+ "kiwi1": [ 0.04908600077033042908, 0.07206099480390548706, 0.04909799993038177490 ],
+ "pear2": [ 0.06601099669933319092, 0.1287339925765991211, 0.06739201396703720093 ],
+ "strawberry1": [0.01698100194334983826, 0.02203200198709964752, 0.01685700193047523499],
+ "apricot": [0.04213499650359153748, 0.05482299625873565674, 0.04333199933171272278],
+ "orange2": [ 0.07349500805139541626, 0.07585700601339340210, 0.07458199560642242432 ],
+ "peach1": [ 0.07397901266813278198, 0.07111301273107528687, 0.07657301425933837891 ],
+ "lemon2": [0.04686100035905838013, 0.04684200137853622437, 0.07244800776243209839],
+ "apple2": [0.05203099921345710754, 0.04766000062227249146, 0.05089000239968299866]}
+
+ for categories in list_categories:
+ point_cloud = f"Models/{categories}/{categories.lower()}.ply"
+ pcd = o3d.io.read_point_cloud(point_cloud)
+
+ fps_points = apply_fps(pcd, 8)
+
+ np.savetxt(f'{dataset_name}/Generated/{categories}/{categories}_fps_3d.txt', fps_points)
+
+ point_cloud_in_numpy = np.asarray(pcd.points)
+ dim = calc_pts_diameter2(point_cloud_in_numpy) * 100
+ np.savetxt(f'{dataset_name}/Generated/{categories}/{categories}_diameter.txt', np.array([dim]))
+
+ size_bb = objs[categories]
+ ext = [x / 2 for x in size_bb]
+ bbox = get_3D_bbox(ext)
+ np.savetxt(f'{dataset_name}/Generated/{categories}/{categories}_bbox_3d.txt', bbox) # save
+
+ #process_compute(dataset_name, camera, new_camera, new_size, Nb_camera, args.World_begin, args.Nb_worlds, list_categories, occ_target, False)
+
diff --git a/prepare_data.py b/prepare_data.py
index 8830d80b2bb81de78255b57eb7ea04cdd77ea24c..4fded51df59985d9d1ebf9e6b5aff7c94386d45e 100644
--- a/prepare_data.py
+++ b/prepare_data.py
@@ -8,6 +8,7 @@ def reform_data(src, data_name, data_option, Nb_camera, World_begin, Nb_world):
shutil.copy(src_meta, dst_meta)
for j in range(1, Nb_camera+1): # cameras
count = ((i-1)*Nb_camera) + j
+ print(count)
if data_option == "ground_truth_rgb":
files_img = os.listdir(f"{src}/{i}/grabber_{j}/color/image/")
src_img = f"{src}/{i}/grabber_{j}/color/image/{files_img[0]}"
diff --git a/slurm/process_dataset.slurm b/slurm/process_dataset.slurm
new file mode 100644
index 0000000000000000000000000000000000000000..65a93b9d7c889b59781ee381bb74c47ad040a1cf
--- /dev/null
+++ b/slurm/process_dataset.slurm
@@ -0,0 +1,40 @@
+#!/bin/bash
+#SBATCH --job-name=Software # nom du job
+#SBATCH -A uli@cpu
+#SBATCH --ntasks=1 # Nombre total de processus MPI
+#SBATCH --ntasks-per-node=1 # Nombre de processus MPI par noeud
+#SBATCH --cpus-per-task=2
+#SBATCH --partition=cpu_p1
+#SBATCH --qos=qos_cpu-t4
+# /!\ Attention, la ligne suivante est trompeuse mais dans le vocabulaire
+# de Slurm "multithread" fait bien référence à l'hyperthreading.
+#SBATCH --hint=nomultithread # 1 processus MPI par coeur physique (pas d'hyperthreading)
+#SBATCH --time=35:00:00 # Temps d’exécution maximum demande (HH:MM:SS)
+#SBATCH --output=Output/postprocess%j.out # Nom du fichier de sortie
+#SBATCH --error=Output/postprocess%j.out # Nom du fichier d'erreur (ici commun avec la sortie)
+
+# on se place dans le répertoire de soumission
+cd ${SLURM_SUBMIT_DIR}
+
+# nettoyage des modules charges en interactif et herites par defaut
+module purge
+
+# chargement des modules
+module load python/3.7.10
+
+# echo des commandes lancées
+set -x
+
+# exécution du code depuis l'espace d’exécution autorisé.
+# Selon l’implémentation de MPI installée dans le conteneur, on positionnera l'option --mpi=pmix_v2 ou --mpi=pmix_v3
+
+id_begin=$1
+Nb=$2
+id_dataset=$3
+rearrange=$4
+compute=$5
+
+conda activate stvNet
+
+time(python main.py --World_begin="$id_begin" --Nb_world="$Nb" --dataset_id="$id_dataset" --rearrange=$rearrange --compute=$compute)
+