Xiaowen/OSAR

Statistics of scenes

Scene type	Train images	Test images	Total	Test ratio
Scene simple	692	88	780	11.28%
Scene complex 1	237	30	267	11.24%
Scene complex 2	111	14	125	11.2%
Total	1040	132	1172	11.26%

Statistics of individual objects

Scene type	Train objects	Test objects	Total	Test ratio
Scene simple	3940	497	4437	11.20%
Scene complex 1	2079	257	2336	11.00%
Scene complex 2	862	111	973	11.41%
Total	6881	865	7746	11.17%

Statistics of expressions

Task	Train samples	Test samples	Total	Test ratio
Object state identification	13760	1730	15490	11.17%
Object grasp prediction	8800	1111	9911	11.21%
Total	22560	2841	25401	11.18%

Statistics of conversation instances

Task	Train samples	Test samples	Total	Test ratio
Object state identification	41280	5190	46470	11.17%
Object grasp prediction	26400	3333	29733	11.21%
Total	67680	8523	76203	11.18%