-
Notifications
You must be signed in to change notification settings - Fork 7
/
get_bbox.py
279 lines (223 loc) · 12.2 KB
/
get_bbox.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
import numpy as np
import PIL
from urdformer import URDFormer
import json
import torch
import cv2
import pybullet as p
from utils import visualization_global, visualization_parts, detection_config
import torchvision.transforms as transforms
from PIL import Image
from scipy.spatial.transform import Rotation as Rot
import argparse
from utils import write_numpy
from utils import write_urdfs
from grounding_dino.detection import detector
from grounding_dino.post_processing import post_processing, summary_kitchen
import os
import time
import glob
import tkinter as tk
from tkinter import filedialog, simpledialog
from PIL import Image, ImageTk
from labeller import BoundingBoxApp
# integrate the extracted texture map into URDFormer prediction
def evaluate_real_image(image_tensor, bbox, masks, tgt_padding_mask, tgt_padding_relation_mask, urdformer, device):
rgb_input = image_tensor.float().to(device).unsqueeze(0)
bbox_input = torch.tensor(bbox).float().to(device).unsqueeze(0)
masks_input = torch.tensor(masks).float().to(device).unsqueeze(0)
tgt_padding_mask = torch.logical_not(tgt_padding_mask)
tgt_padding_mask = torch.tensor(tgt_padding_mask).to(device).unsqueeze(0)
tgt_padding_relation_mask = torch.logical_not(tgt_padding_relation_mask)
tgt_padding_relation_mask = torch.tensor(tgt_padding_relation_mask).to(device).unsqueeze(0)
position_x_pred, position_y_pred, position_z_pred, position_x_end_pred, position_y_end_pred, position_z_end_pred, mesh_pred, parent_cls, base_pred = urdformer(rgb_input, bbox_input, masks_input, 2)
position_pred_x = position_x_pred[tgt_padding_mask].argmax(dim=1)
position_pred_y = position_y_pred[tgt_padding_mask].argmax(dim=1)
position_pred_z = position_z_pred[tgt_padding_mask].argmax(dim=1)
position_pred_x_end = position_x_end_pred[tgt_padding_mask].argmax(dim=1)
position_pred_y_end = position_y_end_pred[tgt_padding_mask].argmax(dim=1)
position_pred_z_end = position_z_end_pred[tgt_padding_mask].argmax(dim=1)
mesh_pred = mesh_pred[tgt_padding_mask].argmax(dim=1)
base_pred = base_pred.argmax(dim=1)
parent_pred = parent_cls[tgt_padding_relation_mask]
position_pred = torch.stack([position_pred_x, position_pred_y, position_pred_z]).T
position_pred_end = torch.stack([position_pred_x_end, position_pred_y_end, position_pred_z_end]).T
return position_pred.detach().cpu().numpy(), position_pred_end.detach().cpu().numpy(), mesh_pred.detach().cpu().numpy(), parent_pred.detach().cpu().numpy(), base_pred.detach().cpu().numpy()
def image_transform():
"""Constructs the image preprocessing transform object.
Arguments:
image_size (int): Size of the result image
"""
# ImageNet normalization statistics
normalize = transforms.Normalize(
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]
)
preprocessing = transforms.Compose([
transforms.Resize(224),
transforms.ToTensor(),
normalize,
])
return preprocessing
def evaluate_parts_with_masks(data_path, cropped_image):
max_bbox = 32
num_roots = 1
data = np.load(data_path, allow_pickle=True).item()
img_pil = PIL.Image.fromarray(cropped_image).resize((224, 224))
img_transform = image_transform()
image_tensor = img_transform(img_pil)
bbox = []
resized_mask = []
for boxid, each_bbox in enumerate(data['part_normalized_bbox']):
bbox.append(each_bbox)
resized = np.zeros((14, 14))
resized_mask.append(resized)
padded_bbox = np.zeros((max_bbox, 4))
padded_bbox[:len(bbox)] = bbox
padded_masks = np.zeros((max_bbox, 14, 14))
padded_masks[:len(resized_mask)] = resized_mask
tgt_padding_mask = torch.ones([max_bbox])
tgt_padding_mask[:len(bbox)] = 0.0
tgt_padding_mask = tgt_padding_mask.bool()
tgt_padding_relation_mask = torch.ones([max_bbox + num_roots])
tgt_padding_relation_mask[:len(bbox) + num_roots] = 0.0
tgt_padding_relation_mask = tgt_padding_relation_mask.bool()
return image_tensor, np.array([padded_bbox]), np.array([padded_masks]), tgt_padding_mask, tgt_padding_relation_mask
def get_binary_relation(global_relations, position_pred_global, num_roots):
new_relations = np.zeros((len(position_pred_global) + num_roots, len(position_pred_global) + num_roots, 6))
for obj_id, position in enumerate(position_pred_global):
each_parent = np.unravel_index(np.argmax(global_relations[num_roots + obj_id]),
global_relations[num_roots + obj_id].shape)
parent_id = each_parent[0]
relation_id = each_parent[1]
new_relations[obj_id + num_roots, parent_id, relation_id] = 1
return new_relations
def get_binary_relation_parts(part_relations, position_pred_part, num_roots):
all_new_relations = []
for obj_id, each_position in enumerate(position_pred_part):
new_relations = np.zeros((len(position_pred_part[obj_id]) + num_roots, len(position_pred_part[obj_id]) + num_roots, 6))
for part_id, position in enumerate(position_pred_part[obj_id]):
part_relations[obj_id][num_roots + part_id][num_roots + part_id] = -1000000000*np.ones(6)# the parent of the one can't be itself...if so, go to the next one.
each_parent = np.unravel_index(np.argmax(part_relations[obj_id][num_roots + part_id]), part_relations[obj_id][num_roots + part_id].shape)
parent_id = each_parent[0]
relation_id = each_parent[1]
new_relations[part_id + num_roots, parent_id, relation_id] = 1
all_new_relations.append(new_relations)
return all_new_relations
def process_gt(gt_data):
part_meshes = gt_data['part_meshes']
part_positions_starts = gt_data['part_positions_start']
part_positions_ends = gt_data['part_positions_end']
new_part_relations = gt_data['part_relations']
base_pred = gt_data['part_bases']
new_starts = []
new_ends = []
for part_id, each_gt_start in enumerate(part_positions_starts):
new_gt_start = [0, each_gt_start[0], each_gt_start[1]]
new_gt_end = [0, part_positions_ends[part_id][0], part_positions_ends[part_id][1]]
new_starts.append(new_gt_start)
new_ends.append(new_gt_end)
new_data = {}
new_data['part_meshes'] = [np.array(part_meshes)]
new_data['part_positions_start'] = [np.array(new_starts)]
new_data['part_positions_end'] = [np.array(new_ends)]
new_data['part_relations'] = [np.array(new_part_relations)]
new_data['part_bases'] = [np.array(base_pred)]
return new_data
def animate():
time.sleep(0.2)
def process_prediction(part_meshes, part_positions_starts, part_positions_ends, part_relations, base_pred):
new_part_relations = get_binary_relation_parts(part_relations, part_positions_starts, 1)
pred_data = {}
if np.array(base_pred)[0] not in [1,2,3,4,5,7]: # if its not cabinet, shelf, oven, dishwasher, washer and fridge, count as rigid
part_meshes = []
part_positions_starts = []
part_positions_ends = []
new_part_relations = np.zeros((1,1, 6))
else:
part_meshes = np.array(part_meshes)[0]
part_positions_starts = np.array(part_positions_starts)[0]
part_positions_ends = np.array(part_positions_ends)[0]
new_part_relations = np.array(new_part_relations)[0]
pred_data['part_meshes'] = [part_meshes]
pred_data['part_positions_start'] = [part_positions_starts]
pred_data['part_positions_end'] = [part_positions_ends]
pred_data['part_relations'] = [new_part_relations]
pred_data['part_bases'] = [np.array(base_pred)[0]]
return pred_data
def evaluate(args, detection_args):
input_path = args.image_path
print('************ Applying Finetuned (Model Soup) GroundingDINO *******************')
detector(args.scene_type, detection_args)
# #
# # # run postprocessing
label_dir = 'grounding_dino/labels'
save_dir = 'grounding_dino/labels_filtered'
manual_dir = 'grounding_dino/labels_manual'
post_processing(label_dir, input_path, save_dir)
# # ask user to for manual correction
#
for img_path in glob.glob(f"{input_path}/*"):
label_name = os.path.basename(img_path)[:-4]
label_path = f"grounding_dino/labels_filtered/{label_name}.npy"
labeled_boxes = np.load(label_path, allow_pickle=True).item()
normalized_bboxes = labeled_boxes['part_normalized_bbox']
root = tk.Tk()
app = BoundingBoxApp(root, img_path, initial_boxes=normalized_bboxes, save_path = manual_dir)
root.mainloop()
if args.scene_type=="kitchen":
detection_args = detection_config(args)
# # get the part detection
os.makedirs(f"{manual_dir}/parts/images", exist_ok=True)
os.makedirs(f"{manual_dir}/parts/labels", exist_ok=True)
os.makedirs(f"{manual_dir}/parts/labels_filtered", exist_ok=True)
os.makedirs(f"{manual_dir}/parts/labels_manual", exist_ok=True)
for each_img_path in glob.glob('images/*'):
label_name = os.path.basename(each_img_path)[:-4]
each_global_label = f"grounding_dino/labels_manual/{label_name}.npy"
global_data = np.load(each_global_label, allow_pickle=True).item()
all_bboxes = global_data['part_normalized_bbox']
image_global = np.array(Image.open(each_img_path).convert("RGB"))
for bbox_id, each_obj_bbox in enumerate(all_bboxes):
bounding_box = [int(each_obj_bbox[0] * image_global.shape[0]),
int(each_obj_bbox[1] * image_global.shape[1]),
int((each_obj_bbox[0] + each_obj_bbox[2]) * image_global.shape[0]),
int((each_obj_bbox[1] + each_obj_bbox[3]) * image_global.shape[1]),
]
cropped_image = image_global[bounding_box[0]:bounding_box[2], bounding_box[1]:bounding_box[3]]
# save the croppd_images in to the folder and update input
PIL.Image.fromarray(cropped_image).resize((512, 512)).save(f"{manual_dir}/parts/images/{label_name}_{bbox_id}.png")
# run detection module again for each cropped images to get part bboxes
detection_args['out_dir'] = "grounding_dino/labels_manual/parts/labels"
detection_args['inputs'] = "grounding_dino/labels_manual/parts/images"
detector('object', detection_args)
#
# # run postprocessing
part_label_dir = 'grounding_dino/labels_manual/parts/labels'
part_save_dir = 'grounding_dino/labels_manual/parts/labels_filtered'
part_manual_dir = 'grounding_dino/labels_manual/parts/labels_manual'
post_processing(part_label_dir, "grounding_dino/labels_manual/parts/images", part_save_dir)
# ask user for manual correction
for img_path in glob.glob("grounding_dino/labels_manual/parts/images/*"):
label_name = os.path.basename(img_path)[:-4]
label_path = f"{part_save_dir}/{label_name}.npy"
labeled_boxes = np.load(label_path, allow_pickle=True).item()
normalized_bboxes = labeled_boxes['part_normalized_bbox']
root = tk.Tk()
app = BoundingBoxApp(root, img_path, initial_boxes=normalized_bboxes, save_path = part_manual_dir)
root.mainloop()
save_dir = f'{manual_dir}/all'
summary_kitchen(manual_dir, part_manual_dir, "images", save_dir)
def main():
parser = argparse.ArgumentParser()
parser.add_argument('-scene_type', '--scene_type', default='cabinet', type=str)
parser.add_argument('-image_path', '--image_path', default='images', type=str)
##################### IMPORTANT! ###############################
# URDFormer replies on good bounding boxes of parts and ojects, you can achieve this by our annotation tool (~1min label per image)
# We also provided our finetuned GroundingDINO (model soup version) to automate this. We finetuned GroundingDino on our generated dataset, and
# apply model soup for the pretrained and finetuned GroundingDINO. However, the perfect bbox prediction is not gauranteed and will be our future work.
args = parser.parse_args()
detection_args = detection_config(args) # leave the defult groundingDINO argument unchanged
evaluate(args, detection_args)
if __name__ == "__main__":
main()