Research Article

Pedestrian Behavior Recognition Based on Improved Dual-stream Network with Differential Feature in Surveillance Video

Table 3

Performance analysis of different comparison models

CategoriesIDTC3DSCNN-3GL-LSTMTs-3DOFGFOur
PrMrRrPrMrRrPrMrRrPrMrRrPrMrRrPrMrRrPrMrRr

Walk652871672273692473752078772875801781791181
Stand732176772078772075832088852176851786862187
Climb513762543264623567672668723762772582791782
Jog67297169247671247168297772297177207987991
Jump613366623269683568613375733366732378721781
Punch425145424144475249524159615165623165682370
Lying573761583267603466573768713168672370681771
Wave 1663266693069693169663277732476761281831286
Wave 269297071317272247769298879298082188789989
Crouch423042443546452447423059542151612362692772
Sitting712579742981732980722582792082811689831588
Handclap383439393543393134383451462459682369732477
Push424745454847434348424758673165722475721780
Fight543658593159573259543668683068641678811481
Handshake633068663271672771633077722178762088772382
Hug682770672873622975772975752779792680821586
Mixed583261603164613064613171702870731978781681