File size: 23,023 Bytes
36c1e62 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 |
import torch
import os
from enum import Enum
from tqdm import tqdm
import numpy as np
from detectron2.structures import BitMasks
from objectrelator.constants import IGNORE_INDEX, IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN, DEFAULT_IM_START_TOKEN, \
DEFAULT_IM_END_TOKEN, DEFAULT_SEG_TOKEN, SEG_TOKEN_INDEX
from objectrelator.model.builder import load_pretrained_model
from objectrelator.utils import disable_torch_init
from objectrelator.mm_utils import tokenizer_image_token, get_model_name_from_path, KeywordsStoppingCriteria
import cv2
from torch.utils.data import Dataset, DataLoader
from objectrelator import conversation as conversation_lib
from objectrelator.train.train_datasets import COCO_interactive_dataset_eval, COCO_interactive_dataset_train, COCO_interactive_dataset
from detectron2.structures import BoxMode
from detectron2.data import MetadataCatalog, DatasetCatalog
from typing import Dict, Optional, Sequence, List
from dataclasses import dataclass, field
import torch.distributed as dist
import transformers
from pathlib import Path
from objectrelator.eval.segmentation_evaluation import openseg_classes
COLOR_MAP = openseg_classes.ADE20K_150_CATEGORIES
import re
from objectrelator.constants import IGNORE_INDEX, IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_PATCH_TOKEN, DEFAULT_IM_START_TOKEN, DEFAULT_IM_END_TOKEN, SEG_TOKEN_INDEX, CLS_TOKEN_INDEX, REGION_TOKEN_INDEX, REFER_TOKEN_INDEX
'''
Create custom dataset classes for EgoExo and Handal datasets
'''
class EgoExo_Dataset_eval(COCO_interactive_dataset_eval):
def preprocess_referring_instruction(self,instruction, REFER_token='[SEG]'):
tokenized = self.tokenizer.encode(instruction, add_special_tokens=False)
tokenized = tokenized + [self.tokenizer.encode(REFER_token, add_special_tokens=False)[0]]
token_refer_id = torch.tensor(tokenized)
return token_refer_id
def tokenizer_special_tokens(self, prompt, tokenizer, image_token_index=IMAGE_TOKEN_INDEX,
seg_token_index=SEG_TOKEN_INDEX, cls_token_index=CLS_TOKEN_INDEX,
region_token_index=REGION_TOKEN_INDEX,refer_token_index=REFER_TOKEN_INDEX, return_tensors=None):
input_ids = []
special_token_map = {'<image>': image_token_index, '<seg>': seg_token_index, '<cls>': cls_token_index, '<region>':region_token_index, '<refer>':refer_token_index}
prompt_chunks = re.split('(<image>|<seg>|<cls>|<region>|<refer>)', prompt)
for chunk in prompt_chunks:
if chunk in special_token_map:
input_ids.append(special_token_map[chunk])
else:
input_ids.extend(tokenizer.encode(chunk, add_special_tokens=False))
if return_tensors is not None:
if return_tensors == 'pt':
return torch.tensor(input_ids, dtype=torch.long).squeeze()
raise ValueError(f'Unsupported tensor type: {return_tensors}')
else:
return input_ids
def __getitem__(self, idx):
data = self.data[idx]
image_file = data['image']
image_folder = self.data_args.image_folder
data_dict = {}
data_dict['file_name'] = os.path.join(image_folder, image_file)
data_dict['height'] = data['image_info']['height']
data_dict['width'] = data['image_info']['width']
data_dict['image_id'] = data['new_img_id']
data_dict['annotations'] = data['anns']
data_dict['vp_annotations'] = data['first_frame_anns']
data_dict['vp_image'] = os.path.join(image_folder,data['first_frame_image'])
for annotation in data_dict['annotations']:
annotation['bbox_mode'] = BoxMode.XYXY_ABS
annotation['bbox'] = [0,0,0,0]
annotation['image_id'] = data['new_img_id']
for annotation in data_dict['vp_annotations']:
annotation['bbox_mode'] = BoxMode.XYXY_ABS
annotation['bbox'] = [0,0,0,0]
annotation['image_id'] = data['new_img_id']
if isinstance(self.data_args.image_processor,dict):
processor = self.data_args.image_processor['instance']
else:
processor = self.data_args.image_processor
region_mask_type = getattr(self.data_args,'region_mask_type',None)
if region_mask_type is not None:
region_mask_type = region_mask_type.split('||')
data_dict = processor.preprocess(data_dict,region_mask_type=region_mask_type,mask_format='bitmask')
sentences = data['instruction']
num_target = len(data_dict['instances'])
prefix_inst = 'This is an image <image>, Please segment by given regions and instruction'
instruction = ''
for sent in sentences:
instruction += ' {}.'.format(sent['sent'])
regions_inst = ' <region>,' * (num_target - 1) + ' <region>.'
sources_value = f'\nThis is all regions: {regions_inst}\n'
sources = [[{'from': 'human', 'value': prefix_inst + sources_value + "and this is the instruction: " + '<refer>\n'},
{'from': 'gpt', 'value': '\n[SEG]<seg>'}]]
text_dict = self.preprocess_llama2(sources, self.tokenizer)
input_ids = text_dict['input_ids'][0]
labels = text_dict['labels'][0]
token_refer_id = self.preprocess_referring_instruction(instruction)
refer_embedding_indices = torch.zeros_like(input_ids)
refer_embedding_indices[input_ids == REFER_TOKEN_INDEX] = 1
data_dict['input_ids'] = input_ids
data_dict['labels'] = labels
data_dict['dataset_type'] = 'referring_coco'
data_dict['token_refer_id'] = token_refer_id
data_dict['refer_embedding_indices'] = refer_embedding_indices
return data_dict
class EgoExo_Dataset_train(COCO_interactive_dataset_train):
def preprocess_referring_instruction(self,instruction, REFER_token='[SEG]'):
tokenized = self.tokenizer.encode(instruction, add_special_tokens=False)
tokenized = tokenized + [self.tokenizer.encode(REFER_token, add_special_tokens=False)[0]]
token_refer_id = torch.tensor(tokenized)
return token_refer_id
def tokenizer_special_tokens(self, prompt, tokenizer, image_token_index=IMAGE_TOKEN_INDEX,
seg_token_index=SEG_TOKEN_INDEX, cls_token_index=CLS_TOKEN_INDEX,
region_token_index=REGION_TOKEN_INDEX,refer_token_index=REFER_TOKEN_INDEX, return_tensors=None):
input_ids = []
special_token_map = {'<image>': image_token_index, '<seg>': seg_token_index, '<cls>': cls_token_index, '<region>':region_token_index, '<refer>':refer_token_index}
prompt_chunks = re.split('(<image>|<seg>|<cls>|<region>|<refer>)', prompt)
for chunk in prompt_chunks:
if chunk in special_token_map:
input_ids.append(special_token_map[chunk])
else:
input_ids.extend(tokenizer.encode(chunk, add_special_tokens=False))
if return_tensors is not None:
if return_tensors == 'pt':
return torch.tensor(input_ids, dtype=torch.long).squeeze()
raise ValueError(f'Unsupported tensor type: {return_tensors}')
else:
return input_ids
def __getitem__(self, idx):
data = self.data[idx]
image_file = data['image']
image_folder = self.data_args.image_folder
data_dict = {}
data_dict['file_name'] = os.path.join(image_folder, image_file)
data_dict['height'] = data['image_info']['height']
data_dict['width'] = data['image_info']['width']
data_dict['image_id'] = data['new_img_id']
data_dict['annotations'] = data['anns']
data_dict['vp_annotations'] = data['first_frame_anns']
data_dict['vp_image'] = os.path.join(image_folder,data['first_frame_image'])
for annotation in data_dict['annotations']:
annotation['bbox_mode'] = BoxMode.XYXY_ABS
annotation['bbox'] = [0,0,0,0]
annotation['image_id'] = data['new_img_id']
for annotation in data_dict['vp_annotations']:
annotation['bbox_mode'] = BoxMode.XYXY_ABS
annotation['bbox'] = [0,0,0,0]
annotation['image_id'] = data['new_img_id']
if isinstance(self.data_args.image_processor,dict):
processor = self.data_args.image_processor['instance']
else:
processor = self.data_args.image_processor
region_mask_type = getattr(self.data_args,'region_mask_type',None)
if region_mask_type is not None:
region_mask_type = region_mask_type.split('||')
data_dict = processor.preprocess(data_dict,region_mask_type=region_mask_type,mask_format='bitmask')
sentences = data['instruction']
num_target = len(data_dict['instances'])
prefix_inst = 'This is an image <image>, Please segment by given regions and instruction'
instruction = ''
for sent in sentences:
instruction += ' {}.'.format(sent['sent'])
regions_inst = ' <region>,' * (num_target - 1) + ' <region>.'
sources_value = f'\nThis is all regions: {regions_inst}\n'
sources = [[{'from': 'human', 'value': prefix_inst + sources_value + "and this is the instruction: " + '<refer>\n'},
{'from': 'gpt', 'value': '\n[SEG]<seg>'}]]
text_dict = self.preprocess_llama2(sources, self.tokenizer)
input_ids = text_dict['input_ids'][0]
labels = text_dict['labels'][0]
token_refer_id = self.preprocess_referring_instruction(instruction)
refer_embedding_indices = torch.zeros_like(input_ids)
refer_embedding_indices[input_ids == REFER_TOKEN_INDEX] = 1
data_dict['input_ids'] = input_ids
data_dict['labels'] = labels
data_dict['dataset_type'] = 'referring_coco'
data_dict['token_refer_id'] = token_refer_id
data_dict['refer_embedding_indices'] = refer_embedding_indices
return data_dict
class Handal_Dataset_eval(COCO_interactive_dataset):
def preprocess_referring_instruction(self,instruction, REFER_token='[SEG]'):
tokenized = self.tokenizer.encode(instruction, add_special_tokens=False)
tokenized = tokenized + [self.tokenizer.encode(REFER_token, add_special_tokens=False)[0]]
token_refer_id = torch.tensor(tokenized)
return token_refer_id
def tokenizer_special_tokens(self, prompt, tokenizer, image_token_index=IMAGE_TOKEN_INDEX,
seg_token_index=SEG_TOKEN_INDEX, cls_token_index=CLS_TOKEN_INDEX,
region_token_index=REGION_TOKEN_INDEX,refer_token_index=REFER_TOKEN_INDEX, return_tensors=None):
input_ids = []
special_token_map = {'<image>': image_token_index, '<seg>': seg_token_index, '<cls>': cls_token_index, '<region>':region_token_index, '<refer>':refer_token_index}
prompt_chunks = re.split('(<image>|<seg>|<cls>|<region>|<refer>)', prompt)
for chunk in prompt_chunks:
if chunk in special_token_map:
input_ids.append(special_token_map[chunk])
else:
input_ids.extend(tokenizer.encode(chunk, add_special_tokens=False))
if return_tensors is not None:
if return_tensors == 'pt':
return torch.tensor(input_ids, dtype=torch.long).squeeze()
raise ValueError(f'Unsupported tensor type: {return_tensors}')
else:
return input_ids
def __getitem__(self, idx):
data = self.data[idx]
image_file = data['image']
image_folder = self.data_args.image_folder
data_dict = {}
data_dict['file_name'] = os.path.join(image_folder, image_file)
data_dict['height'] = data['image_info']['height']
data_dict['width'] = data['image_info']['width']
data_dict['image_id'] = data['new_img_id']
data_dict['annotations'] = data['anns']
data_dict['vp_annotations'] = data['first_frame_anns']
data_dict['vp_image'] = os.path.join(image_folder,data['first_frame_image'])
for annotation in data_dict['annotations']:
annotation['bbox_mode'] = BoxMode.XYXY_ABS
annotation['bbox'] = [0,0,0,0]
annotation['image_id'] = data['new_img_id']
for annotation in data_dict['vp_annotations']:
annotation['bbox_mode'] = BoxMode.XYXY_ABS
annotation['bbox'] = [0,0,0,0]
annotation['image_id'] = data['new_img_id']
if isinstance(self.data_args.image_processor,dict):
processor = self.data_args.image_processor['instance']
else:
processor = self.data_args.image_processor
region_mask_type = getattr(self.data_args,'region_mask_type',None)
if region_mask_type is not None:
region_mask_type = region_mask_type.split('||')
data_dict = processor.preprocess(data_dict,region_mask_type=region_mask_type,mask_format='bitmask')
sentences = data['instruction']
num_target = len(data_dict['instances'])
prefix_inst = 'This is an image <image>, Please segment by given regions and instruction'
instruction = ''
for sent in sentences:
instruction += ' {}.'.format(sent['sent'])
regions_inst = ' <region>,' * (num_target - 1) + ' <region>.'
sources_value = f'\nThis is all regions: {regions_inst}\n'
sources = [[{'from': 'human', 'value': prefix_inst + sources_value + "and this is the instruction: " + '<refer>\n'},
{'from': 'gpt', 'value': '\n[SEG]<seg>'}]]
text_dict = self.preprocess_llama2(sources, self.tokenizer)
input_ids = text_dict['input_ids'][0]
labels = text_dict['labels'][0]
token_refer_id = self.preprocess_referring_instruction(instruction)
refer_embedding_indices = torch.zeros_like(input_ids)
refer_embedding_indices[input_ids == REFER_TOKEN_INDEX] = 1
data_dict['input_ids'] = input_ids
data_dict['labels'] = labels
data_dict['dataset_type'] = 'referring_coco'
data_dict['token_refer_id'] = token_refer_id
data_dict['refer_embedding_indices'] = refer_embedding_indices
return data_dict
class Handal_Dataset_train(COCO_interactive_dataset_train):
def preprocess_referring_instruction(self,instruction, REFER_token='[SEG]'):
tokenized = self.tokenizer.encode(instruction, add_special_tokens=False)
tokenized = tokenized + [self.tokenizer.encode(REFER_token, add_special_tokens=False)[0]]
token_refer_id = torch.tensor(tokenized)
return token_refer_id
def tokenizer_special_tokens(self, prompt, tokenizer, image_token_index=IMAGE_TOKEN_INDEX,
seg_token_index=SEG_TOKEN_INDEX, cls_token_index=CLS_TOKEN_INDEX,
region_token_index=REGION_TOKEN_INDEX,refer_token_index=REFER_TOKEN_INDEX, return_tensors=None):
input_ids = []
special_token_map = {'<image>': image_token_index, '<seg>': seg_token_index, '<cls>': cls_token_index, '<region>':region_token_index, '<refer>':refer_token_index}
prompt_chunks = re.split('(<image>|<seg>|<cls>|<region>|<refer>)', prompt)
for chunk in prompt_chunks:
if chunk in special_token_map:
input_ids.append(special_token_map[chunk])
else:
input_ids.extend(tokenizer.encode(chunk, add_special_tokens=False))
if return_tensors is not None:
if return_tensors == 'pt':
return torch.tensor(input_ids, dtype=torch.long).squeeze()
raise ValueError(f'Unsupported tensor type: {return_tensors}')
else:
return input_ids
def __getitem__(self, idx):
data = self.data[idx]
image_file = data['image']
image_folder = self.data_args.image_folder
data_dict = {}
data_dict['file_name'] = os.path.join(image_folder, image_file)
data_dict['height'] = data['image_info']['height']
data_dict['width'] = data['image_info']['width']
data_dict['image_id'] = data['new_img_id']
data_dict['annotations'] = data['anns']
data_dict['vp_annotations'] = data['first_frame_anns']
data_dict['vp_image'] = os.path.join(image_folder,data['first_frame_image'])
for annotation in data_dict['annotations']:
annotation['bbox_mode'] = BoxMode.XYXY_ABS
annotation['bbox'] = [0,0,0,0]
annotation['image_id'] = data['new_img_id']
for annotation in data_dict['vp_annotations']:
annotation['bbox_mode'] = BoxMode.XYXY_ABS
annotation['bbox'] = [0,0,0,0]
annotation['image_id'] = data['new_img_id']
if isinstance(self.data_args.image_processor,dict):
processor = self.data_args.image_processor['instance']
else:
processor = self.data_args.image_processor
region_mask_type = getattr(self.data_args,'region_mask_type',None)
if region_mask_type is not None:
region_mask_type = region_mask_type.split('||')
data_dict = processor.preprocess(data_dict,region_mask_type=region_mask_type,mask_format='bitmask')
sentences = data['instruction']
num_target = len(data_dict['instances'])
prefix_inst = 'This is an image <image>, Please segment by given regions and instruction'
instruction = ''
for sent in sentences:
instruction += ' {}.'.format(sent['sent'])
regions_inst = ' <region>,' * (num_target - 1) + ' <region>.'
sources_value = f'\nThis is all regions: {regions_inst}\n'
sources = [[{'from': 'human', 'value': prefix_inst + sources_value + "and this is the instruction: " + '<refer>\n'},
{'from': 'gpt', 'value': '\n[SEG]<seg>'}]]
text_dict = self.preprocess_llama2(sources, self.tokenizer)
input_ids = text_dict['input_ids'][0]
labels = text_dict['labels'][0]
token_refer_id = self.preprocess_referring_instruction(instruction)
refer_embedding_indices = torch.zeros_like(input_ids)
refer_embedding_indices[input_ids == REFER_TOKEN_INDEX] = 1
data_dict['input_ids'] = input_ids
data_dict['labels'] = labels
data_dict['dataset_type'] = 'referring_coco'
data_dict['token_refer_id'] = token_refer_id
data_dict['refer_embedding_indices'] = refer_embedding_indices
return data_dict
class EgoExo_Dataset_eval_crossview_refseg(COCO_interactive_dataset_eval):
def preprocess_referring_instruction(self,instruction, REFER_token='[SEG]'):
tokenized = self.tokenizer.encode(instruction, add_special_tokens=False)
tokenized = tokenized + [self.tokenizer.encode(REFER_token, add_special_tokens=False)[0]]
token_refer_id = torch.tensor(tokenized)
return token_refer_id
def tokenizer_special_tokens(self, prompt, tokenizer, image_token_index=IMAGE_TOKEN_INDEX,
seg_token_index=SEG_TOKEN_INDEX, cls_token_index=CLS_TOKEN_INDEX,
region_token_index=REGION_TOKEN_INDEX,refer_token_index=REFER_TOKEN_INDEX, return_tensors=None):
input_ids = []
special_token_map = {'<image>': image_token_index, '<seg>': seg_token_index, '<cls>': cls_token_index, '<region>':region_token_index, '<refer>':refer_token_index}
prompt_chunks = re.split('(<image>|<seg>|<cls>|<region>|<refer>)', prompt)
for chunk in prompt_chunks:
if chunk in special_token_map:
input_ids.append(special_token_map[chunk])
else:
input_ids.extend(tokenizer.encode(chunk, add_special_tokens=False))
if return_tensors is not None:
if return_tensors == 'pt':
return torch.tensor(input_ids, dtype=torch.long).squeeze()
raise ValueError(f'Unsupported tensor type: {return_tensors}')
else:
return input_ids
def __getitem__(self, idx):
data = self.data[idx]
image_file = data['image']
image_folder = self.data_args.image_folder
data_dict = {}
data_dict['file_name'] = os.path.join(image_folder, image_file)
data_dict['height'] = data['image_info']['height']
data_dict['width'] = data['image_info']['width']
data_dict['image_id'] = data['new_img_id']
data_dict['annotations'] = data['anns']
data_dict['vp_annotations'] = data['first_frame_anns']
data_dict['vp_image'] = os.path.join(image_folder,data['first_frame_image'])
for annotation in data_dict['annotations']:
annotation['bbox_mode'] = BoxMode.XYXY_ABS
annotation['bbox'] = [0,0,0,0]
annotation['image_id'] = data['new_img_id']
for annotation in data_dict['vp_annotations']:
annotation['bbox_mode'] = BoxMode.XYXY_ABS
annotation['bbox'] = [0,0,0,0]
annotation['image_id'] = data['new_img_id']
if isinstance(self.data_args.image_processor,dict):
processor = self.data_args.image_processor['crossview_refseg']
else:
processor = self.data_args.image_processor
region_mask_type = getattr(self.data_args,'region_mask_type',None)
if region_mask_type is not None:
region_mask_type = region_mask_type.split('||')
data_dict = processor.preprocess(data_dict,region_mask_type=region_mask_type,mask_format='bitmask')
sentences = data['instruction']
num_target = len(data_dict['instances'])
prefix_inst = 'This is an image <image>, Please segment by given regions and instruction'
instruction = ''
for sent in sentences:
instruction += ' {}.'.format(sent['sent'])
regions_inst = ' <region>,' * (num_target - 1) + ' <region>.'
sources_value = f'\nThis is all regions: {regions_inst}\n'
sources = [[{'from': 'human', 'value': prefix_inst + sources_value + "and this is the instruction: " + '<refer>\n'},
{'from': 'gpt', 'value': '\n[SEG]<seg>'}]]
text_dict = self.preprocess_llama2(sources, self.tokenizer)
input_ids = text_dict['input_ids'][0]
labels = text_dict['labels'][0]
token_refer_id = self.preprocess_referring_instruction(instruction)
refer_embedding_indices = torch.zeros_like(input_ids)
refer_embedding_indices[input_ids == REFER_TOKEN_INDEX] = 1
data_dict['input_ids'] = input_ids
data_dict['labels'] = labels
data_dict['dataset_type'] = 'referring_coco'
data_dict['token_refer_id'] = token_refer_id
data_dict['refer_embedding_indices'] = refer_embedding_indices
return data_dict |