AnimeHeadsv3 / README.md
nyuuzyou
added dataset without augmentation, changed the script
aa35aae
|
raw
history blame
2.72 kB
metadata
task_categories:
  - object-detection
license: wtfpl
dataset_info:
  - config_name: With augmentation
    features:
      - name: image_id
        dtype: int64
      - name: image
        dtype: image
      - name: width
        dtype: int32
      - name: height
        dtype: int32
      - name: objects
        sequence:
          - name: id
            dtype: int64
          - name: area
            dtype: int64
          - name: bbox
            sequence: float32
            length: 4
          - name: category
            dtype: string
    splits:
      - name: train
        num_bytes: 2817954
        num_examples: 8037
      - name: validation
        num_bytes: 37647
        num_examples: 100
      - name: test
        num_bytes: 8425
        num_examples: 20
    download_size: 590150250
    dataset_size: 2864026
  - config_name: Without augmentation
    features:
      - name: image_id
        dtype: int64
      - name: image
        dtype: image
      - name: width
        dtype: int32
      - name: height
        dtype: int32
      - name: objects
        sequence:
          - name: id
            dtype: int64
          - name: area
            dtype: int64
          - name: bbox
            sequence: float32
            length: 4
          - name: category
            dtype: string
    splits:
      - name: train
        num_bytes: 932413
        num_examples: 2659
      - name: validation
        num_bytes: 37647
        num_examples: 100
      - name: test
        num_bytes: 7393
        num_examples: 18
    download_size: 512953012
    dataset_size: 977453

AnimeHeadsv3 Object Detection Dataset

The AnimeHeadsv3 Object Detection Dataset is a collection of anime and art images, including manga pages, that have been annotated with object bounding boxes for use in object detection tasks.

Contents

There are two versions of the dataset available: The dataset contains a total of 8157 images, split into training, validation, and testing sets. The images were collected from various sources and include a variety of anime and art styles, including manga.

  • Dataset with augmentation: Contains 8157 images.
  • Dataset without augmentation: Contains 2777 images.

The images were collected from various sources and include a variety of anime and art styles, including manga. The annotations were created using the COCO format, with each annotation file containing the bounding box coordinates and label for each object in the corresponding image. The dataset has only one class named "head".

Preprocessing

The dataset with augmentation has the following preprocessing parameters:

Resize: Fit within 640x640

The dataset without augmentation does not have any preprocessing applied.

Augmentation Parameters

The following augmentation parameters were applied to the dataset with augmentation:

Outputs per training example: 3
Flip: Horizontal
Saturation: Between -40% and +40%
Blur: Up to 4px
Noise: Up to 4% of pixels