Datasets:

ClarusC64
/

false_absence_detection_v01

sample_id stringlengths 10 10	split stringclasses 3 values	modality stringclasses 1 value	scene_type stringclasses 5 values	sequence_id stringclasses 5 values	frame_index int64 0 32	container_id stringclasses 5 values	container_bounds stringclasses 3 values	boundary_type stringclasses 2 values	zone_id stringlengths 6 6	zone_type stringlengths 4 12	target_entity_id stringclasses 5 values	target_visibility stringclasses 2 values	absence_tag stringclasses 3 values	evidence_type stringclasses 9 values	false_absence_target bool 2 classes	trap_flag bool 2 classes	label_type stringclasses 3 values	drift_risk stringclasses 3 values	comment stringlengths 13 61
fad_000001	train	video	indoor_room	seq_0001	0	room_01	0 0 1920 1080	hard	zone_A	sofa	cat_01	visible	present	none	false	false	baseline	low	cat on sofa fully visible
fad_000002	train	video	indoor_room	seq_0001	1	room_01	0 0 1920 1080	hard	zone_A	sofa	cat_01	not_visible	still_present	shadow	true	true	false_absence_case	medium	cat drops behind sofa, tail shadow still visible
fad_000003	train	video	indoor_room	seq_0001	2	room_01	0 0 1920 1080	hard	zone_B	door	cat_01	not_visible	left_scene	door_state	false	false	true_exit	medium	door opens and closes as cat leaves room
fad_000004	train	video	factory_line	seq_0002	10	line_01	0 0 2560 1440	hard	zone_C	conveyor	crate_05	visible	present	none	false	false	baseline	low	crate on belt with clear view
fad_000005	train	video	factory_line	seq_0002	11	line_01	0 0 2560 1440	hard	zone_C	conveyor	crate_05	not_visible	still_present	motion_trail	true	true	false_absence_case	high	belt motion and timing imply crate still on line
fad_000006	train	video	factory_line	seq_0002	12	line_01	0 0 2560 1440	hard	zone_D	chute	crate_05	not_visible	left_scene	interaction_trace	false	false	true_exit	high	crate heard dropping into chute bin
fad_000007	train	video	corridor	seq_0003	4	corr_01	0 0 1080 720	soft	zone_E	passage	cart_02	visible	present	none	false	false	baseline	low	cart mid corridor
fad_000008	train	video	corridor	seq_0003	5	corr_01	0 0 1080 720	soft	zone_F	turn_left	cart_02	not_visible	still_present	audio_cue	true	true	false_absence_case	medium	wheel noise continues around corner
fad_000009	train	video	corridor	seq_0003	6	corr_01	0 0 1080 720	soft	zone_G	exit_door	cart_02	not_visible	left_scene	door_state	false	false	true_exit	medium	exit door opens, wheel noise stops
fad_000010	valid	video	sports_pitch	seq_0100	20	pitch_01	0 0 1920 1080	soft	zone_L	left_flank	ball_07	visible	present	none	false	false	baseline	low	ball near touchline
fad_000011	valid	video	sports_pitch	seq_0100	21	pitch_01	0 0 1920 1080	soft	zone_L	left_flank	ball_07	not_visible	still_present	reflection	true	true	false_absence_case	high	ball hidden behind player, reflection on ad board
fad_000012	valid	video	sports_pitch	seq_0100	22	pitch_01	0 0 1920 1080	soft	zone_M	stands	ball_07	not_visible	left_scene	crowd_reaction	false	false	true_exit	high	crowd reacts as ball enters stands
fad_000013	eval	video	warehouse	seq_0201	30	wh_02	0 0 2560 1440	hard	zone_P	robot_lane	robot_11	visible	present	none	false	false	baseline	low	robot in lane
fad_000014	eval	video	warehouse	seq_0201	31	wh_02	0 0 2560 1440	hard	zone_Q	intersection	robot_11	not_visible	still_present	physical_constraint	true	true	false_absence_case	high	layout makes teleport away impossible, robot blocked by racks
fad_000015	eval	video	warehouse	seq_0201	32	wh_02	0 0 2560 1440	hard	zone_R	dock_exit	robot_11	not_visible	left_scene	none	false	false	true_exit	medium	later frame shows empty lane toward dock, robot assumed gone

ClarusC64/false_absence_detection_v01

Dataset summary

This dataset tests whether models treat invisible entities as gone or still present.
In each sequence, a target leaves the camera view.
Some exits are real.
Some are false absences with clear evidence that the target remains in the container.

Goal

check if models infer continued presence from indirect cues
avoid treating every disappearance as an exit
keep spatial grounding under occlusion and clutter

Key signals

absence_tag: present, still_present, left_scene
evidence_type: shadow, reflection, audio cue, motion trail, interaction trace, door state, physical constraint, none
false_absence_target: true when the model should infer “still present”
trap_flag: true when the sample is designed to lure models into assuming exit

Columns

sample_id – unique id per frame sample
split – train, valid, eval
modality – video
scene_type – indoor_room, factory_line, corridor, sports_pitch, warehouse
sequence_id – id for a temporal sequence
frame_index – index within the sequence
time_gap – not used here (fixed gaps inside raw video)
container_id – id of the main container
container_bounds – "x_min y_min x_max y_max"
boundary_type – hard, soft, porous
zone_id – region id inside the container
zone_type – sofa, door, conveyor, chute, passage, turn_left, exit_door, left_flank, stands, robot_lane, intersection, dock_exit
target_entity_id – tracked entity such as cat_01, crate_05, cart_02, ball_07, robot_11
target_visibility – visible, partial, not_visible
absence_tag – present, still_present, left_scene
evidence_type – none, shadow, motion_trail, interaction_trace, audio_cue, door_state, reflection, crowd_reaction, physical_constraint
false_absence_target – true if the model should infer continued presence from evidence
trap_flag – true if the scene is designed as a false-absence trap
label_type – baseline, false_absence_case, true_exit
drift_risk – low, medium, high
comment – short human note

Example loading code

from datasets import load_dataset

ds = load_dataset("ClarusC64/false_absence_detection_v01")

row = ds["train"][1]
print(row["target_entity_id"], row["target_visibility"],
      row["absence_tag"], row["evidence_type"], row["false_absence_target"])

Downloads last month: 6

Size of downloaded dataset files:

3.24 kB

Size of the auto-converted Parquet files:

10.3 kB

Number of rows: