Spaces:
Sleeping
Sleeping
File size: 2,052 Bytes
c6827c1 |
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 |
import torch
import numpy as np
from torchvision import transforms as T
from torch.utils.data import Dataset
from PIL import Image
from torchvision.ops.boxes import box_convert
import glob
import json
# dataset
# |_ train
# | |_ .BMP files
# | |_ annotation
# | |_ classes.txt (one class per line)
# | |_ .txt anno files (class x_center y_center width height)
# |_ test
# |_ val
class Therin(Dataset): # Therin: Intruder thermal dataset
def __init__(self, dir, set):
self._dir = dir + '/' + set
# self._imglist = glob.glob(self._dir + '/*.BMP')
self._json_path = dir + '/' + set + '.json'
with open(self._json_path) as anno_file:
self._anno = json.load(anno_file)["annotations"]
with open(self._json_path) as anno_file:
self._imglist = json.load(anno_file)["images"]
self._transform = T.Compose([T.ToTensor()])
def __len__(self):
return len(self._imglist)
def __getitem__(self, index):
image = Image.open(self._dir + "/" + self._imglist[index]["file_name"])
boxes = np.zeros((1, 4), dtype=np.float32)
boxes[0] = self._anno[index]['bbox']
boxes = torch.as_tensor(boxes, dtype=torch.float32)
boxes = box_convert(boxes, in_fmt='xywh', out_fmt='xyxy')
gt_classes = np.zeros((1), dtype=np.int32)
gt_classes[0] = self._anno[index]['category_id']-1
gt_classes = torch.as_tensor(gt_classes, dtype=torch.int64)
image_id = self._anno[index]['image_id']
image_id = torch.as_tensor(image_id, dtype=torch.int64)
area = np.zeros((1), dtype=np.int32)
area[0] = self._anno[index]['area']
area = torch.as_tensor(area, dtype=torch.int64)
target = {"labels": gt_classes, "boxes": boxes, "image_id": image_id, "area": area, "iscrowd": torch.tensor([0])}
image = self._transform(image)
return image, target
def collate_fn(self, batch):
return tuple(zip(*batch)) |