metadata
dataset_info:
- config_name: task_1_volume
features:
- name: sample_id
dtype: string
- name: query_image_1_color
dtype: image
- name: query_image_2_color
dtype: image
- name: query_image_1_color_axes
dtype: image
- name: query_image_2_color_axes
dtype: image
- name: query_image_1_uncolor
dtype: image
- name: query_image_2_uncolor
dtype: image
- name: query_image_1_uncolor_axes
dtype: image
- name: query_image_2_uncolor_axes
dtype: image
- name: answer
dtype: int32
- name: distractors
list: int32
- name: base_blocks_json
dtype: string
splits:
- name: train
num_bytes: 48676689
num_examples: 300
download_size: 48723789
dataset_size: 48676689
- config_name: task_2_packaging
features:
- name: sample_id
dtype: string
- name: query_image_1_color
dtype: image
- name: query_image_2_color
dtype: image
- name: query_image_1_color_axes
dtype: image
- name: query_image_2_color_axes
dtype: image
- name: query_image_1_uncolor
dtype: image
- name: query_image_2_uncolor
dtype: image
- name: query_image_1_uncolor_axes
dtype: image
- name: query_image_2_uncolor_axes
dtype: image
- name: box_dims
list: int32
- name: answer
dtype: bool
splits:
- name: train
num_bytes: 48633092
num_examples: 300
download_size: 48707929
dataset_size: 48633092
- config_name: task_3_difference
features:
- name: pair_id
dtype: string
- name: sample_a_id
dtype: string
- name: sample_b_id
dtype: string
- name: query_a_image_1_color
dtype: image
- name: query_a_image_2_color
dtype: image
- name: query_a_image_1_color_axes
dtype: image
- name: query_a_image_2_color_axes
dtype: image
- name: query_a_image_1_uncolor
dtype: image
- name: query_a_image_2_uncolor
dtype: image
- name: query_a_image_1_uncolor_axes
dtype: image
- name: query_a_image_2_uncolor_axes
dtype: image
- name: query_b_image_1_color
dtype: image
- name: query_b_image_2_color
dtype: image
- name: query_b_image_1_color_axes
dtype: image
- name: query_b_image_2_color_axes
dtype: image
- name: query_b_image_1_uncolor
dtype: image
- name: query_b_image_2_uncolor
dtype: image
- name: query_b_image_1_uncolor_axes
dtype: image
- name: query_b_image_2_uncolor_axes
dtype: image
- name: vol_a
dtype: int32
- name: vol_b
dtype: int32
- name: answer_intersection
dtype: int32
- name: answer_only_a
dtype: int32
- name: answer_only_b
dtype: int32
- name: answer_color_aware_intersection
dtype: int32
- name: answer_color_aware_only_a
dtype: int32
- name: answer_color_aware_only_b
dtype: int32
splits:
- name: train
num_bytes: 97413442
num_examples: 300
download_size: 97542358
dataset_size: 97413442
- config_name: task_4_composition
features:
- name: sample_id
dtype: string
- name: query_image_1_color
dtype: image
- name: query_image_2_color
dtype: image
- name: query_image_1_color_axes
dtype: image
- name: query_image_2_color_axes
dtype: image
- name: color_distribution_json
dtype: string
- name: n_distinct_colors
dtype: int32
splits:
- name: train
num_bytes: 25770903
num_examples: 300
download_size: 25781755
dataset_size: 25770903
- config_name: task_a_projection
features:
- name: sample_id
dtype: string
- name: subtask
dtype: string
- name: query_image_1_color
dtype: image
- name: query_image_2_color
dtype: image
- name: query_image_1_color_axes
dtype: image
- name: query_image_2_color_axes
dtype: image
- name: query_image_1_uncolor
dtype: image
- name: query_image_2_uncolor
dtype: image
- name: query_image_1_uncolor_axes
dtype: image
- name: query_image_2_uncolor_axes
dtype: image
- name: candidate_0_color
dtype: image
- name: candidate_1_color
dtype: image
- name: candidate_2_color
dtype: image
- name: candidate_0_uncolor
dtype: image
- name: candidate_1_uncolor
dtype: image
- name: candidate_2_uncolor
dtype: image
- name: answer_idx
dtype: int32
- name: answer_plane
dtype: string
splits:
- name: train
num_bytes: 311777317
num_examples: 1788
download_size: 313986512
dataset_size: 311777317
- config_name: task_b_reconstruct
features:
- name: sample_id
dtype: string
- name: distractor_sample_id
dtype: string
- name: silhouette_0_color
dtype: image
- name: silhouette_1_color
dtype: image
- name: silhouette_2_color
dtype: image
- name: silhouette_0_uncolor
dtype: image
- name: silhouette_1_uncolor
dtype: image
- name: silhouette_2_uncolor
dtype: image
- name: candidate_0_color
dtype: image
- name: candidate_0_color_axes
dtype: image
- name: candidate_0_uncolor
dtype: image
- name: candidate_0_uncolor_axes
dtype: image
- name: candidate_1_color
dtype: image
- name: candidate_1_color_axes
dtype: image
- name: candidate_1_uncolor
dtype: image
- name: candidate_1_uncolor_axes
dtype: image
- name: answer_idx
dtype: int32
splits:
- name: train
num_bytes: 52934910
num_examples: 300
download_size: 53088010
dataset_size: 52934910
configs:
- config_name: task_1_volume
data_files:
- split: train
path: task_1_volume/train-*
- config_name: task_2_packaging
data_files:
- split: train
path: task_2_packaging/train-*
- config_name: task_3_difference
data_files:
- split: train
path: task_3_difference/train-*
- config_name: task_4_composition
data_files:
- split: train
path: task_4_composition/train-*
- config_name: task_a_projection
data_files:
- split: train
path: task_a_projection/train-*
- config_name: task_b_reconstruct
data_files:
- split: train
path: task_b_reconstruct/train-*
BED: Blind 3D (B3D) dataset for VLMs
- Main Authors: Nguyen Kim Hai Bui, An Ngo Xuan
- Initial Idea: Khai Le Duc