The recognition rate of the YOLOv2 example in the example is very low.

[zx310]

I trained a YOLOv2 model using MaxHub, and after training, the recognition rate was 0.955, which is quite high. However, why does the example produce so many boxes—no fewer than 254? This is really strange. Could it be that I modified something incorrectly in the example?

./yolov2
MaixHub detection yolov2
================================ model stat ================================
mdl_type=0 (int8))
out_deq=1
input_cnt=1, output_cnt=1, layer_cnt=28
input 3dims: (224, 224, 3)
output 3dims: (7, 7, 30)
main buf size 903168; sub buf size 0
//Note: PARAM is layer param size, include align padding

Idx Layer outshape inoft outoft PARAM MEMOUT OPS
--- Input 224,224, 3 - 0 0 150528 0
###L71: body oft = 64
###L76: type=0, is_out=0, size=920, in_oft=0, out_oft=602112, in_dims=[3,224,224,3], out_dims=[3,112,112,24], in_s=0.017, in_zp=-5, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=3, kh=3, sw=2, sh=2, dw=1, dh=1, act=3, pad=[0,1,0,1], dmul=0, ws_oft=80, w_oft=176, b_oft=824
000 Conv2D 112,112, 24 0 602112 840 301056 8128512
###L71: body oft = 984
###L76: type=5, is_out=0, size=488, in_oft=602112, out_oft=0, in_dims=[3,112,112,24], out_dims=[3,112,112,24], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=1, sh=1, dw=1, dh=1, act=3, pad=[1,1,1,1], dmul=1, ws_oft=80, w_oft=176, b_oft=392
001 DWConv2D 112,112, 24 602112 0 408 301056 2709504
###L71: body oft = 1472
###L76: type=0, is_out=0, size=1616, in_oft=0, out_oft=301056, in_dims=[3,112,112,24], out_dims=[3,112,112,48], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=272, b_oft=1424
002 Conv2D 112,112, 48 0 301056 1536 602112 14450688
###L71: body oft = 3088
###L76: type=5, is_out=0, size=896, in_oft=301056, out_oft=0, in_dims=[3,112,112,48], out_dims=[3,56,56,48], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=2, sh=2, dw=1, dh=1, act=3, pad=[0,1,0,1], dmul=1, ws_oft=80, w_oft=272, b_oft=704
003 DWConv2D 56, 56, 48 301056 0 816 150528 1354752
###L71: body oft = 3984
###L76: type=0, is_out=0, size=5456, in_oft=0, out_oft=602112, in_dims=[3,56,56,48], out_dims=[3,56,56,96], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=464, b_oft=5072
004 Conv2D 56, 56, 96 0 602112 5376 301056 14450688
###L71: body oft = 9440
###L76: type=5, is_out=0, size=1712, in_oft=602112, out_oft=0, in_dims=[3,56,56,96], out_dims=[3,56,56,96], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=1, sh=1, dw=1, dh=1, act=3, pad=[1,1,1,1], dmul=1, ws_oft=80, w_oft=464, b_oft=1328
005 DWConv2D 56, 56, 96 602112 0 1632 301056 2709504
###L71: body oft = 11152
###L76: type=0, is_out=0, size=10064, in_oft=0, out_oft=602112, in_dims=[3,56,56,96], out_dims=[3,56,56,96], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=464, b_oft=9680
006 Conv2D 56, 56, 96 0 602112 9984 301056 28901376
###L71: body oft = 21216
###L76: type=5, is_out=0, size=1712, in_oft=602112, out_oft=0, in_dims=[3,56,56,96], out_dims=[3,28,28,96], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=2, sh=2, dw=1, dh=1, act=3, pad=[0,1,0,1], dmul=1, ws_oft=80, w_oft=464, b_oft=1328
007 DWConv2D 28, 28, 96 602112 0 1632 75264 677376
###L71: body oft = 22928
###L76: type=0, is_out=0, size=20048, in_oft=0, out_oft=752640, in_dims=[3,28,28,96], out_dims=[3,28,28,192], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=848, b_oft=19280
008 Conv2D 28, 28,192 0 752640 19968 150528 14450688
###L71: body oft = 42976
###L76: type=5, is_out=0, size=3344, in_oft=752640, out_oft=0, in_dims=[3,28,28,192], out_dims=[3,28,28,192], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=1, sh=1, dw=1, dh=1, act=3, pad=[1,1,1,1], dmul=1, ws_oft=80, w_oft=848, b_oft=2576
009 DWConv2D 28, 28,192 752640 0 3264 150528 1354752
###L71: body oft = 46320
###L76: type=0, is_out=0, size=38480, in_oft=0, out_oft=752640, in_dims=[3,28,28,192], out_dims=[3,28,28,192], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=848, b_oft=37712
010 Conv2D 28, 28,192 0 752640 38400 150528 28901376
###L71: body oft = 84800
###L76: type=5, is_out=0, size=3344, in_oft=752640, out_oft=0, in_dims=[3,28,28,192], out_dims=[3,14,14,192], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=2, sh=2, dw=1, dh=1, act=3, pad=[0,1,0,1], dmul=1, ws_oft=80, w_oft=848, b_oft=2576
011 DWConv2D 14, 14,192 752640 0 3264 37632 338688
###L71: body oft = 88144
###L76: type=0, is_out=0, size=76880, in_oft=0, out_oft=827904, in_dims=[3,14,14,192], out_dims=[3,14,14,384], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=1616, b_oft=75344
012 Conv2D 14, 14,384 0 827904 76800 75264 14450688
###L71: body oft = 165024
###L76: type=5, is_out=0, size=6608, in_oft=827904, out_oft=0, in_dims=[3,14,14,384], out_dims=[3,14,14,384], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=1, sh=1, dw=1, dh=1, act=3, pad=[1,1,1,1], dmul=1, ws_oft=80, w_oft=1616, b_oft=5072
013 DWConv2D 14, 14,384 827904 0 6528 75264 677376
###L71: body oft = 171632
###L76: type=0, is_out=0, size=150608, in_oft=0, out_oft=827904, in_dims=[3,14,14,384], out_dims=[3,14,14,384], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=1616, b_oft=149072
014 Conv2D 14, 14,384 0 827904 150528 75264 28901376
###L71: body oft = 322240
###L76: type=5, is_out=0, size=6608, in_oft=827904, out_oft=0, in_dims=[3,14,14,384], out_dims=[3,14,14,384], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=1, sh=1, dw=1, dh=1, act=3, pad=[1,1,1,1], dmul=1, ws_oft=80, w_oft=1616, b_oft=5072
015 DWConv2D 14, 14,384 827904 0 6528 75264 677376
###L71: body oft = 328848
###L76: type=0, is_out=0, size=150608, in_oft=0, out_oft=827904, in_dims=[3,14,14,384], out_dims=[3,14,14,384], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=1616, b_oft=149072
016 Conv2D 14, 14,384 0 827904 150528 75264 28901376
###L71: body oft = 479456
###L76: type=5, is_out=0, size=6608, in_oft=827904, out_oft=0, in_dims=[3,14,14,384], out_dims=[3,14,14,384], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=1, sh=1, dw=1, dh=1, act=3, pad=[1,1,1,1], dmul=1, ws_oft=80, w_oft=1616, b_oft=5072
017 DWConv2D 14, 14,384 827904 0 6528 75264 677376
###L71: body oft = 486064
###L76: type=0, is_out=0, size=150608, in_oft=0, out_oft=827904, in_dims=[3,14,14,384], out_dims=[3,14,14,384], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=1616, b_oft=149072
018 Conv2D 14, 14,384 0 827904 150528 75264 28901376
###L71: body oft = 636672
###L76: type=5, is_out=0, size=6608, in_oft=827904, out_oft=0, in_dims=[3,14,14,384], out_dims=[3,14,14,384], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=1, sh=1, dw=1, dh=1, act=3, pad=[1,1,1,1], dmul=1, ws_oft=80, w_oft=1616, b_oft=5072
019 DWConv2D 14, 14,384 827904 0 6528 75264 677376
###L71: body oft = 643280
###L76: type=0, is_out=0, size=150608, in_oft=0, out_oft=827904, in_dims=[3,14,14,384], out_dims=[3,14,14,384], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=1616, b_oft=149072
020 Conv2D 14, 14,384 0 827904 150528 75264 28901376
###L71: body oft = 793888
###L76: type=5, is_out=0, size=6608, in_oft=827904, out_oft=0, in_dims=[3,14,14,384], out_dims=[3,14,14,384], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=1, sh=1, dw=1, dh=1, act=3, pad=[1,1,1,1], dmul=1, ws_oft=80, w_oft=1616, b_oft=5072
021 DWConv2D 14, 14,384 827904 0 6528 75264 677376
###L71: body oft = 800496
###L76: type=0, is_out=0, size=150608, in_oft=0, out_oft=827904, in_dims=[3,14,14,384], out_dims=[3,14,14,384], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=1616, b_oft=149072
022 Conv2D 14, 14,384 0 827904 150528 75264 28901376
###L71: body oft = 951104
###L76: type=5, is_out=0, size=6608, in_oft=827904, out_oft=0, in_dims=[3,14,14,384], out_dims=[3,7,7,384], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=2, sh=2, dw=1, dh=1, act=3, pad=[0,1,0,1], dmul=1, ws_oft=80, w_oft=1616, b_oft=5072
023 DWConv2D 7, 7,384 827904 0 6528 18816 169344
###L71: body oft = 957712
###L76: type=0, is_out=0, size=301136, in_oft=0, out_oft=865536, in_dims=[3,7,7,384], out_dims=[3,7,7,768], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=3152, b_oft=298064
024 Conv2D 7, 7,768 0 865536 301056 37632 14450688
###L71: body oft = 1258848
###L76: type=5, is_out=0, size=13136, in_oft=865536, out_oft=0, in_dims=[3,7,7,768], out_dims=[3,7,7,768], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L108: DWConv2d: kw=3, kh=3, sw=1, sh=1, dw=1, dh=1, act=3, pad=[1,1,1,1], dmul=1, ws_oft=80, w_oft=3152, b_oft=10064
025 DWConv2D 7, 7,768 865536 0 13056 37632 338688
###L71: body oft = 1271984
###L76: type=0, is_out=0, size=596048, in_oft=0, out_oft=865536, in_dims=[3,7,7,768], out_dims=[3,7,7,768], in_s=0.024, in_zp=-128, out_s=0.024, out_zp=-128
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=3, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=3152, b_oft=592976
026 Conv2D 7, 7,768 0 865536 595968 37632 28901376
###L71: body oft = 1868032
###L76: type=0, is_out=1, size=23360, in_oft=865536, out_oft=0, in_dims=[3,7,7,768], out_dims=[3,7,7,30], in_s=0.024, in_zp=-128, out_s=0.137, out_zp=96
###L88: Conv2d: kw=1, kh=1, sw=1, sh=1, dw=1, dh=1, act=0, pad=[0,0,0,0], dmul=0, ws_oft=80, w_oft=200, b_oft=23240
027 Conv2D 7, 7, 30 865536 0 23280 1470 1128960

Total param ~1844.8 KB, OPS ~325.76 MOPS, buffer 882.0 KB

===tm_run use 410.438 ms
decode result boxes_num: 245
----image:0----

img:0, box: x:-19, y:0, w:118, h:46, class:0, prob:0.972, label:bus

img:0, box: x:132, y:16, w:78, h:23, class:0, prob:0.964, label:bus

img:0, box: x:134, y:59, w:136, h:40, class:0, prob:0.976, label:bus

img:0, box: x:-38, y:105, w:156, h:30, class:0, prob:0.871, label:bus

img:0, box: x:18, y:120, w:39, h:46, class:0, prob:0.953, label:bus

img:0, box: x:-25, y:192, w:60, h:40, class:0, prob:0.964, label:bus

img:0, box: x:38, y:199, w:60, h:35, class:0, prob:0.979, label:bus

img:0, box: x:108, y:194, w:52, h:26, class:0, prob:0.981, label:bus

img:0, box: x:110, y:71, w:11, h:54, class:0, prob:0.976, label:bus

img:0, box: x:21, y:144, w:15, h:41, class:0, prob:0.984, label:bus

img:0, box: x:180, y:139, w:10, h:47, class:0, prob:0.871, label:bus

img:0, box: x:172, y:174, w:10, h:47, class:0, prob:0.930, label:bus

img:0, box: x:208, y:175, w:10, h:41, class:0, prob:0.953, label:bus

img:0, box: x:-101, y:-21, w:241, h:89, class:0, prob:0.959, label:bus

img:0, box: x:71, y:-48, w:210, h:118, class:0, prob:0.981, label:bus

img:0, box: x:-4, y:22, w:364, h:68, class:0, prob:0.964, label:bus

img:0, box: x:107, y:2, w:184, h:177, class:0, prob:0.855, label:bus

img:0, box: x:-98, y:69, w:478, h:78, class:0, prob:0.946, label:bus

img:0, box: x:146, y:79, w:122, h:78, class:0, prob:0.968, label:bus

img:0, box: x:-5, y:115, w:160, h:118, class:0, prob:0.986, label:bus

img:0, box: x:162, y:140, w:92, h:78, class:0, prob:0.939, label:bus

img:0, box: x:-70, y:146, w:364, h:118, class:0, prob:0.979, label:bus

img:0, box: x:127, y:21, w:165, h:1, class:0, prob:0.981, label:bus

img:0, box: x:61, y:147, w:286, h:2, class:0, prob:0.871, label:bus

img:0, box: x:137, y:-57, w:74, h:184, class:0, prob:0.976, label:bus

img:0, box: x:2, y:-9, w:97, h:161, class:0, prob:0.953, label:bus

img:0, box: x:110, y:-16, w:56, h:161, class:0, prob:0.959, label:bus

img:0, box: x:-22, y:81, w:56, h:106, class:0, prob:0.972, label:bus

img:0, box: x:46, y:77, w:49, h:106, class:0, prob:0.930, label:bus

img:0, box: x:84, y:83, w:97, h:93, class:0, prob:0.921, label:bus

img:0, box: x:59, y:10, w:42, h:318, class:0, prob:0.959, label:bus

img:0, box: x:63, y:177, w:74, h:41, class:0, prob:0.986, label:bus