Code: Select all
Block 32 has strict standard length for a commercial.
Block 32 score: Before - 1.00 After - 3.00
Block 32 has logo.
Block 32 score: Before - 3.00 After - 0.03
Block 32 AR (1.74) is different from dominant AR(1.34).
Block 32 score: Before - 0.03 After - 0.06
<snip>
Block list after weighing
----------------------------------------------------
# sbf bs be fs fe sc len scr cmb ar cut bri logo vol sil corr stdev cc
32:++ 6 2 4 157728 161334 0.06 120.320 0.07 0 1.74 A L S ub 407- 0.38 337- 23+ 0.000 40
Code: Select all
Block 43 has logo.
Block 43 score: Before - 1.00 After - 0.01
<snip>
Block list after weighing
----------------------------------------------------
# sbf bs be fs fe sc len scr cmb ar cut bri logo vol sil corr stdev cc
43:++ 2 2 0 212462 217482 0.01 167.501 0.09 0 1.34 L a 676+ 0.27 330- 20- 0.000 67
It looks like both of these are results of false-positive logo detection (there's nothing that looks like logo in these commercials). What's the best way to crack down on the logo detection without losing part of the show?
Thanks