Datasets:
metadata
dataset_info:
features:
- name: height
dtype: int64
- name: width
dtype: int64
- name: image
dtype: image
- name: image_id
dtype: int64
- name: objects
struct:
- name: area
sequence: float64
- name: bbox
sequence:
sequence: float64
- name: catagories
sequence: string
splits:
- name: train
num_bytes: 3366913203
num_examples: 10000
download_size: 3329769639
dataset_size: 3366913203
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
task_categories:
- object-detection
size_categories:
- 1K<n<10K
DAC System Design Contest 2023 Dataset
dataset is in coco format and with images ending in _*.jpg removed. I did not make any splits, in the effort to keep it close to the original.