Function 'load_models' executed in 8.916s
load checkpoint from local path: OBJECT_DETECTOR_FOLDER/mask2former_swin-s-p4-w7-224_lsj_8x2_50e_coco.pth
Summary
=======
Total images: 2212
Total prompts: 553
% correct images: 89.33%
% correct prompts: 89.33%

Task breakdown
==============
two_object       = 97.98% (388 / 396)
counting         = 90.00% (288 / 320)
position         = 82.00% (328 / 400)
color_attr       = 77.00% (308 / 400)
single_object    = 100.00% (320 / 320)
colors           = 91.49% (344 / 376)

Overall score (avg. over tasks): 0.89745
