metadata
license: fair-noncommercial-research-license
configs:
- config_name: mot
data_files:
- split: train
path: mot/train-*
- config_name: obj_det_seg
data_files:
- split: acs_ec
path: obj_det_seg/acs_ec-*
- split: acs_eg
path: obj_det_seg/acs_eg-*
- split: ie_central
path: obj_det_seg/ie_central-*
- split: r_central
path: obj_det_seg/r_central-*
default: true
- config_name: obj_det_seg_efficient_grounded_sam
data_files:
- split: acs_ec
path: obj_det_seg_efficient_grounded_sam/acs_ec-*
- split: acs_eg
path: obj_det_seg_efficient_grounded_sam/acs_eg-*
- split: ie_central
path: obj_det_seg_efficient_grounded_sam/ie_central-*
- split: r_central
path: obj_det_seg_efficient_grounded_sam/r_central-*
- config_name: obj_det_seg_grounded_sam
data_files:
- split: acs_ec
path: obj_det_seg_grounded_sam/acs_ec-*
- split: acs_eg
path: obj_det_seg_grounded_sam/acs_eg-*
- split: ie_central
path: obj_det_seg_grounded_sam/ie_central-*
- split: r_central
path: obj_det_seg_grounded_sam/r_central-*
- config_name: sam3
data_files:
- split: train
path: sam3/train-*
- config_name: tracking
data_files:
- split: acs_ec
path: tracking/acs_ec-*
- split: acs_eg
path: tracking/acs_eg-*
- split: ie_central
path: tracking/ie_central-*
- split: r_central
path: tracking/r_central-*
dataset_info:
- config_name: mot
features:
- name: image
dtype: image
- name: type
dtype: string
- name: scene
dtype: string
- name: objects
struct:
- name: bbox
list:
list: float32
length: 4
- name: category
list:
class_label:
names:
'0': person
- name: area
list: float32
- name: id
list: int64
- name: score
list: float32
- name: track_id
list: int64
- name: image_id
dtype: int64
- name: width
dtype: int32
- name: height
dtype: int32
- name: recording_id
dtype: string
- name: file_name
dtype: string
splits:
- name: train
num_bytes: 317178253
num_examples: 1305
download_size: 309416593
dataset_size: 317178253
- config_name: obj_det_seg
features:
- name: image
dtype: image
- name: type
dtype: string
- name: label_source
dtype: string
- name: scene
dtype: string
- name: objects
struct:
- name: bbox
list:
list: float32
length: 4
- name: category
list:
class_label:
names:
'0': person
- name: area
list: float32
- name: id
list: int64
- name: score
list: float32
- name: track_id
list: int64
- name: image_id
dtype: int64
- name: width
dtype: int32
- name: height
dtype: int32
- name: recording_id
dtype: string
- name: file_name
dtype: string
splits:
- name: acs_ec
num_bytes: 643471147
num_examples: 2802
- name: acs_eg
num_bytes: 630798025
num_examples: 3975
- name: ie_central
num_bytes: 452418319
num_examples: 1650
- name: r_central
num_bytes: 83111291
num_examples: 380
download_size: 1843335951
dataset_size: 1809798782
- config_name: obj_det_seg_efficient_grounded_sam
features:
- name: image
dtype: image
- name: type
dtype: string
- name: label_source
dtype: string
- name: scene
dtype: string
- name: objects
struct:
- name: bbox
list:
list: float32
length: 4
- name: category
list:
class_label:
names:
'0': person
- name: area
list: float32
- name: id
list: int64
- name: score
list: float32
- name: track_id
list: int64
- name: image_id
dtype: int64
- name: width
dtype: int32
- name: height
dtype: int32
- name: recording_id
dtype: string
- name: file_name
dtype: string
splits:
- name: acs_ec
num_bytes: 640472191
num_examples: 2802
- name: acs_eg
num_bytes: 630117991
num_examples: 3975
- name: ie_central
num_bytes: 452026243
num_examples: 1650
- name: r_central
num_bytes: 83092451
num_examples: 380
download_size: 1838936390
dataset_size: 1805708876
- config_name: obj_det_seg_grounded_sam
features:
- name: image
dtype: image
- name: type
dtype: string
- name: label_source
dtype: string
- name: scene
dtype: string
- name: objects
struct:
- name: bbox
list:
list: float32
length: 4
- name: category
list:
class_label:
names:
'0': person
- name: area
list: float32
- name: id
list: int64
- name: score
list: float32
- name: track_id
list: int64
- name: image_id
dtype: int64
- name: width
dtype: int32
- name: height
dtype: int32
- name: recording_id
dtype: string
- name: file_name
dtype: string
splits:
- name: acs_ec
num_bytes: 640444171
num_examples: 2802
- name: acs_eg
num_bytes: 630078241
num_examples: 3975
- name: ie_central
num_bytes: 452009743
num_examples: 1650
- name: r_central
num_bytes: 83088651
num_examples: 380
download_size: 1838935198
dataset_size: 1805620806
- config_name: sam3
features:
- name: image
dtype: image
- name: type
dtype: string
- name: scene
dtype: string
- name: objects
struct:
- name: bbox
list:
list: float32
length: 4
- name: category
list:
class_label:
names:
'0': person
- name: area
list: float32
- name: id
list: int64
- name: score
list: float32
- name: track_id
list: int64
- name: image_id
dtype: int64
- name: width
dtype: int32
- name: height
dtype: int32
- name: recording_id
dtype: string
- name: file_name
dtype: string
splits:
- name: train
num_bytes: 2016686472
num_examples: 8807
download_size: 1843335662
dataset_size: 2016686472
- config_name: tracking
features:
- name: image
dtype: image
- name: type
dtype: string
- name: scene
dtype: string
- name: objects
struct:
- name: bbox
list:
list: float32
length: 4
- name: category
list:
class_label:
names:
'0': person
- name: area
list: float32
- name: id
list: int64
- name: score
list: float32
- name: track_id
list: int64
- name: image_id
dtype: int64
- name: width
dtype: int32
- name: height
dtype: int32
- name: recording_id
dtype: string
- name: file_name
dtype: string
splits:
- name: acs_ec
num_bytes: 42616582
num_examples: 174
- name: acs_eg
num_bytes: 34449048
num_examples: 215
- name: ie_central
num_bytes: 149362435
num_examples: 536
- name: r_central
num_bytes: 83078500
num_examples: 380
download_size: 309433429
dataset_size: 309506565
Dataset Card for IndoorCrowd
Dataset Description
- Paper: IndoorCrowd: A Multi-Scene Dataset for Human Detection, Segmentation, and Tracking with an Automated Annotation Pipeline (CVPR 2026 Submission)
Dataset Summary
IndoorCrowd is a multi-scene dataset designed for indoor human detection, instance segmentation, and multi-object tracking. It captures diverse challenges such as viewpoint variation, partial occlusion, and varying crowd density across four distinct campus locations (ACS-EC, ACS-EG, IE-Central, R-Central). Faces are explicitly blurred to preserve privacy, making it suitable for safe research into intelligent crowd management and behaviour tracking.
The dataset consists of 31 videos sampled at 5 FPS, totalling 9,913 frames.
Subsets
- Object Detection and Segmentation: 9,913 frames featuring bounding boxes and instance segmentation masks. Includes a rigorously annotated 620-frame pure-human control subset for foundation-model benchmarking.
- Multi-Object Tracking (MOT): A 2,552-frame tracking subset providing continuous identity tracks following the MOTChallenge format.
Supported Tasks
object-detection: Detecting human bounding boxes (Baselines benchmarked: YOLOv8n, YOLOv26n, RT-DETR-L).image-segmentation: Generating instance-level masks for people in crowded indoor geometries.video-object-tracking: Maintaining human identity across consecutive frames via tracking algorithms (Baselines benchmarked: ByteTrack, BoT-SORT, OC-SORT).
Dataset Creation
Curation Rationale
Outdoor datasets currently dominate development. Indoor environments introduce a new set of challenges like camera view obstructions (pillars, furniture), structural occlusions, near-to-distal scale variance, and abrupt density fluctuations.
Annotations
Annotations were produced using a semi-automated pipeline:
- Auto-labelling: Uses foundation models such as SAM3, GroundingSAM, and EfficientGroundingSAM to generate initial candidate masks and tracklets.
- Human Correction: Expert human reviewers used SAM 2.1 to manually delete false positives, append missing masks, correct identity switches, and linearly interpolate gaps, ensuring high-fidelity ground truth.
Data Splits
The dataset provides varied crowd density regimes:
- ACS-EC: A dense multi-level atrium setting with small instance scales and high occlusion ($79.3%$ dense frames).
- ACS-EG: A narrow ground-level corridor with substantial person scale variations lengthways.
- IE-Central: An intermediate seating/entrance hall environment.
- R-Central: An overhead-view atrium with prominent structural columns causing regular occlusions.
Personal and Sensitive Information
All human faces in the raw footage have been strictly blurred by an automated de-identification pipeline prior to release. No audio, demographic attributes, or personal identifiers are collected.
Additional Information
Licensing Information
The dataset is released under a license restricting its use strictly to non-commercial computer vision research. It prohibits surveillance and any re-identification of individuals.