metadata

dataset_info:
  features:
    - name: image
      dtype: image
    - name: bounding_boxes
      sequence:
        sequence: float64
    - name: classes
      sequence: int64
  splits:
    - name: train
      num_bytes: 147837671.888
      num_examples: 2596
    - name: validation
      num_bytes: 26070707
      num_examples: 458
  download_size: 172845541
  dataset_size: 173908378.888
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
license: apache-2.0
task_categories:
  - image-classification
pretty_name: FLIR IR YOLO expansion
size_categories:
  - 1K<n<10K

FLIR IR YOLO expansion

Images from repository FLIR_IR_Expansion

This expansion pack is prepared specifically for training a YOU-ONLY-LOOK-ONCE(YOLO) network. All frames are labeled in the YOLO format. If you want to use this expansion pack for other purposes, images are still available for download but requires manual labeling

Download all dataset

If you want to download all dataset you must do

from datasets import load_dataset

dataset = load_dataset("SAE-AAI/FLIR_IR_Expansion")

If you only want to download train or validation split you must do

from datasets import load_dataset

train_dataset = load_dataset("SAE-AAI/FLIR_IR_Expansion", split='train')
validation_dataset = load_dataset("SAE-AAI/FLIR_IR_Expansion", split='validation')

Download by stream

If you want to donwload dataset by stream, you must do

from datasets import load_dataset

iterable_dataset = load_dataset("SAE-AAI/FLIR_IR_Expansion", streaming=True)

now you can get every sample by

for sample in iterable_dataset['train']:
    print(sample['bounding_boxes'], sample['classes'])
    example['sample'].show()
    break

or with

sample = next(iter(iterable_dataset['train']))
print(sample['bounding_boxes'], sample['classes'])
sample['image']

If you want to get a batch of samples you ca do

BS = 4
example_batch = list(iterable_dataset['train'].take(BS))