This data set contains 14 video sequences and ground truth files for stationary foreground segmentation tasks and abandoned & stolen object detection.
Sequence | Source | Stationary objects | |
---|---|---|---|
VISOR_00 | visor_Video00 1:2308 | 720:2308 875:1610 | car #1 car #2 |
VISOR_01 | visor_Video01 1:3330 | 970:2720 | car |
VISOR_02 | visor_Video02 1:2815 | 850:2700 1530:2710 | car #1 car #2 |
VISOR_03 | visor_Video03 1:3075 | 975:2385 | car |
AVSS_00a | AVSSS07_EASY 1500:3775 | 500:2275 | suitcase |
AVSS_00b | AVSSS07_EASY 3780:5465 | 1:1100 | suitcase |
AVSS_01a | AVSSS07_MEDIUM 1:3000 | 1410:3000 | bag |
AVSS_01b | AVSSS07_MEDIUM 2880:4812 | 1:1650 | bag |
PETS_00a | PETS2006_S1_C3 1:2992 | 1890:2992 | bag |
PETS_00b | PETS2006_S1_C3 2992:1 | 1:1190 | bag |
PETS_01a | PETS2006_S4_C3 1:3052 | 840:3052 | suitcase |
PETS_01b | PETS2006_S4_C3 3052:1 | 1:2265 | suitcase |
PETS_02a | PETS2006_S5_C3 1:3361 | 1860:3361 | plank |
PETS_02b | PETS2006_S5_C3 3361:1 | 1:1590 | plank |
Ground truth files are comma separated text files. For every frame containing a stationay object the ground truth file contains a row starting with the frame index (from 1) and then for each blob of the frame the bounding box and class (abandoned / stolen) is given.
E.g. for two blobs
frame;x1,y1,w1,h1,C1;x2,y2,w2,h2,C2;
where frame
is the frame index, the blob descriptions are also spearated with ;
and the blob description gives the x, y coordinates and width, hegith. The label ob the blob can be A
if abandoned or S
if stolen.
The test videos used in this data set are from the VISOR, AVSS2007 and PETS2006 sets.