Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Tiny model works, but Large model fails to predict. #187

Open
XononoX opened this issue Mar 25, 2022 · 0 comments
Open

Tiny model works, but Large model fails to predict. #187

XononoX opened this issue Mar 25, 2022 · 0 comments

Comments

@XononoX
Copy link

XononoX commented Mar 25, 2022 •
edited
Loading

Hello!

I'm using the windows version of darknet to train models in Yolo-v4. I'm specifically using only the CPU for predictions, this program will need to run on machines with only integrated graphics hardware.

As in the title, I've got a tiny model and a large model, both trained on the same datasets. I'm able to load both with darknet.exe for the test operation and get predictions from either one. Of course, the large model performs better, gives more consistent and more accurate bounding boxes, so I'd love to be able to implement the large model in my C# application, but unfortunately, as of right now, while I can load both models with Alturos.Yolo, I can't get predictions as I'm expecting from my large model.

Here are the .cfg files, converted to .txt, for each of the models...
YoloV4_Custom_Large.txt
YoloV4_Custom_Tiny.txt

Here's the test image that I've been trying to feed:
0005_0 5
The file size limit prevents me from uploading the full sized, 5100x7019 pixel version, so I scaled it down by 50%.

Here are some screenshots of the predictions that each model gives on that image in darknet.exe, first the Large model:
Large_Predictions

And the Tiny model:
Tiny_Predictions

Now, when I load up the model in my C# application and try to predict on this same image, it initializes using the same .cfg that I provided and loads the weights, giving the following output to the console, but not finding any bounding boxes:

Initializing YoloV4 AI...
If YOLO fails check for Microsoft Visual C++ 2017 Redistributable (x64) vc_redist.x64.exe
policy: Using default 'constant'
batch = 1, time_steps = 1, train = 0
   layer   filters  size/strd(dil)      input                output
   0 conv     32       3 x 3/ 1    512 x 704 x   3 ->  512 x 704 x  32 0.623 BF
   1 conv     64       3 x 3/ 2    512 x 704 x  32 ->  256 x 352 x  64 3.322 BF
   2 conv     64       1 x 1/ 1    256 x 352 x  64 ->  256 x 352 x  64 0.738 BF
   3 route  1                                      ->  256 x 352 x  64
   4 conv     64       1 x 1/ 1    256 x 352 x  64 ->  256 x 352 x  64 0.738 BF
   5 conv     32       1 x 1/ 1    256 x 352 x  64 ->  256 x 352 x  32 0.369 BF
   6 conv     64       3 x 3/ 1    256 x 352 x  32 ->  256 x 352 x  64 3.322 BF
   7 Shortcut Layer: 4,  wt = 0, wn = 0, outputs: 256 x 352 x  64 0.006 BF
   8 conv     64       1 x 1/ 1    256 x 352 x  64 ->  256 x 352 x  64 0.738 BF
   9 route  8 2                                    ->  256 x 352 x 128
  10 conv     64       1 x 1/ 1    256 x 352 x 128 ->  256 x 352 x  64 1.476 BF
  11 conv    128       3 x 3/ 2    256 x 352 x  64 ->  128 x 176 x 128 3.322 BF
  12 conv     64       1 x 1/ 1    128 x 176 x 128 ->  128 x 176 x  64 0.369 BF
  13 route  11                                     ->  128 x 176 x 128
  14 conv     64       1 x 1/ 1    128 x 176 x 128 ->  128 x 176 x  64 0.369 BF
  15 conv     64       1 x 1/ 1    128 x 176 x  64 ->  128 x 176 x  64 0.185 BF
  16 conv     64       3 x 3/ 1    128 x 176 x  64 ->  128 x 176 x  64 1.661 BF
  17 Shortcut Layer: 14,  wt = 0, wn = 0, outputs: 128 x 176 x  64 0.001 BF
  18 conv     64       1 x 1/ 1    128 x 176 x  64 ->  128 x 176 x  64 0.185 BF
  19 conv     64       3 x 3/ 1    128 x 176 x  64 ->  128 x 176 x  64 1.661 BF
  20 Shortcut Layer: 17,  wt = 0, wn = 0, outputs: 128 x 176 x  64 0.001 BF
  21 conv     64       1 x 1/ 1    128 x 176 x  64 ->  128 x 176 x  64 0.185 BF
  22 route  21 12                                  ->  128 x 176 x 128
  23 conv    128       1 x 1/ 1    128 x 176 x 128 ->  128 x 176 x 128 0.738 BF
  24 conv    256       3 x 3/ 2    128 x 176 x 128 ->   64 x  88 x 256 3.322 BF
  25 conv    128       1 x 1/ 1     64 x  88 x 256 ->   64 x  88 x 128 0.369 BF
  26 route  24                                     ->   64 x  88 x 256
  27 conv    128       1 x 1/ 1     64 x  88 x 256 ->   64 x  88 x 128 0.369 BF
  28 conv    128       1 x 1/ 1     64 x  88 x 128 ->   64 x  88 x 128 0.185 BF
  29 conv    128       3 x 3/ 1     64 x  88 x 128 ->   64 x  88 x 128 1.661 BF
  30 Shortcut Layer: 27,  wt = 0, wn = 0, outputs:  64 x  88 x 128 0.001 BF
  31 conv    128       1 x 1/ 1     64 x  88 x 128 ->   64 x  88 x 128 0.185 BF
  32 conv    128       3 x 3/ 1     64 x  88 x 128 ->   64 x  88 x 128 1.661 BF
  33 Shortcut Layer: 30,  wt = 0, wn = 0, outputs:  64 x  88 x 128 0.001 BF
  34 conv    128       1 x 1/ 1     64 x  88 x 128 ->   64 x  88 x 128 0.185 BF
  35 conv    128       3 x 3/ 1     64 x  88 x 128 ->   64 x  88 x 128 1.661 BF
  36 Shortcut Layer: 33,  wt = 0, wn = 0, outputs:  64 x  88 x 128 0.001 BF
  37 conv    128       1 x 1/ 1     64 x  88 x 128 ->   64 x  88 x 128 0.185 BF
  38 conv    128       3 x 3/ 1     64 x  88 x 128 ->   64 x  88 x 128 1.661 BF
  39 Shortcut Layer: 36,  wt = 0, wn = 0, outputs:  64 x  88 x 128 0.001 BF
  40 conv    128       1 x 1/ 1     64 x  88 x 128 ->   64 x  88 x 128 0.185 BF
  41 conv    128       3 x 3/ 1     64 x  88 x 128 ->   64 x  88 x 128 1.661 BF
  42 Shortcut Layer: 39,  wt = 0, wn = 0, outputs:  64 x  88 x 128 0.001 BF
  43 conv    128       1 x 1/ 1     64 x  88 x 128 ->   64 x  88 x 128 0.185 BF
  44 conv    128       3 x 3/ 1     64 x  88 x 128 ->   64 x  88 x 128 1.661 BF
  45 Shortcut Layer: 42,  wt = 0, wn = 0, outputs:  64 x  88 x 128 0.001 BF
  46 conv    128       1 x 1/ 1     64 x  88 x 128 ->   64 x  88 x 128 0.185 BF
  47 conv    128       3 x 3/ 1     64 x  88 x 128 ->   64 x  88 x 128 1.661 BF
  48 Shortcut Layer: 45,  wt = 0, wn = 0, outputs:  64 x  88 x 128 0.001 BF
  49 conv    128       1 x 1/ 1     64 x  88 x 128 ->   64 x  88 x 128 0.185 BF
  50 conv    128       3 x 3/ 1     64 x  88 x 128 ->   64 x  88 x 128 1.661 BF
  51 Shortcut Layer: 48,  wt = 0, wn = 0, outputs:  64 x  88 x 128 0.001 BF
  52 conv    128       1 x 1/ 1     64 x  88 x 128 ->   64 x  88 x 128 0.185 BF
  53 route  52 25                                  ->   64 x  88 x 256
  54 conv    256       1 x 1/ 1     64 x  88 x 256 ->   64 x  88 x 256 0.738 BF
  55 conv    512       3 x 3/ 2     64 x  88 x 256 ->   32 x  44 x 512 3.322 BF
  56 conv    256       1 x 1/ 1     32 x  44 x 512 ->   32 x  44 x 256 0.369 BF
  57 route  55                                     ->   32 x  44 x 512
  58 conv    256       1 x 1/ 1     32 x  44 x 512 ->   32 x  44 x 256 0.369 BF
  59 conv    256       1 x 1/ 1     32 x  44 x 256 ->   32 x  44 x 256 0.185 BF
  60 conv    256       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 256 1.661 BF
  61 Shortcut Layer: 58,  wt = 0, wn = 0, outputs:  32 x  44 x 256 0.000 BF
  62 conv    256       1 x 1/ 1     32 x  44 x 256 ->   32 x  44 x 256 0.185 BF
  63 conv    256       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 256 1.661 BF
  64 Shortcut Layer: 61,  wt = 0, wn = 0, outputs:  32 x  44 x 256 0.000 BF
  65 conv    256       1 x 1/ 1     32 x  44 x 256 ->   32 x  44 x 256 0.185 BF
  66 conv    256       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 256 1.661 BF
  67 Shortcut Layer: 64,  wt = 0, wn = 0, outputs:  32 x  44 x 256 0.000 BF
  68 conv    256       1 x 1/ 1     32 x  44 x 256 ->   32 x  44 x 256 0.185 BF
  69 conv    256       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 256 1.661 BF
  70 Shortcut Layer: 67,  wt = 0, wn = 0, outputs:  32 x  44 x 256 0.000 BF
  71 conv    256       1 x 1/ 1     32 x  44 x 256 ->   32 x  44 x 256 0.185 BF
  72 conv    256       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 256 1.661 BF
  73 Shortcut Layer: 70,  wt = 0, wn = 0, outputs:  32 x  44 x 256 0.000 BF
  74 conv    256       1 x 1/ 1     32 x  44 x 256 ->   32 x  44 x 256 0.185 BF
  75 conv    256       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 256 1.661 BF
  76 Shortcut Layer: 73,  wt = 0, wn = 0, outputs:  32 x  44 x 256 0.000 BF
  77 conv    256       1 x 1/ 1     32 x  44 x 256 ->   32 x  44 x 256 0.185 BF
  78 conv    256       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 256 1.661 BF
  79 Shortcut Layer: 76,  wt = 0, wn = 0, outputs:  32 x  44 x 256 0.000 BF
  80 conv    256       1 x 1/ 1     32 x  44 x 256 ->   32 x  44 x 256 0.185 BF
  81 conv    256       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 256 1.661 BF
  82 Shortcut Layer: 79,  wt = 0, wn = 0, outputs:  32 x  44 x 256 0.000 BF
  83 conv    256       1 x 1/ 1     32 x  44 x 256 ->   32 x  44 x 256 0.185 BF
  84 route  83 56                                  ->   32 x  44 x 512
  85 conv    512       1 x 1/ 1     32 x  44 x 512 ->   32 x  44 x 512 0.738 BF
  86 conv   1024       3 x 3/ 2     32 x  44 x 512 ->   16 x  22 x1024 3.322 BF
  87 conv    512       1 x 1/ 1     16 x  22 x1024 ->   16 x  22 x 512 0.369 BF
  88 route  86                                     ->   16 x  22 x1024
  89 conv    512       1 x 1/ 1     16 x  22 x1024 ->   16 x  22 x 512 0.369 BF
  90 conv    512       1 x 1/ 1     16 x  22 x 512 ->   16 x  22 x 512 0.185 BF
  91 conv    512       3 x 3/ 1     16 x  22 x 512 ->   16 x  22 x 512 1.661 BF
  92 Shortcut Layer: 89,  wt = 0, wn = 0, outputs:  16 x  22 x 512 0.000 BF
  93 conv    512       1 x 1/ 1     16 x  22 x 512 ->   16 x  22 x 512 0.185 BF
  94 conv    512       3 x 3/ 1     16 x  22 x 512 ->   16 x  22 x 512 1.661 BF
  95 Shortcut Layer: 92,  wt = 0, wn = 0, outputs:  16 x  22 x 512 0.000 BF
  96 conv    512       1 x 1/ 1     16 x  22 x 512 ->   16 x  22 x 512 0.185 BF
  97 conv    512       3 x 3/ 1     16 x  22 x 512 ->   16 x  22 x 512 1.661 BF
  98 Shortcut Layer: 95,  wt = 0, wn = 0, outputs:  16 x  22 x 512 0.000 BF
  99 conv    512       1 x 1/ 1     16 x  22 x 512 ->   16 x  22 x 512 0.185 BF
 100 conv    512       3 x 3/ 1     16 x  22 x 512 ->   16 x  22 x 512 1.661 BF
 101 Shortcut Layer: 98,  wt = 0, wn = 0, outputs:  16 x  22 x 512 0.000 BF
 102 conv    512       1 x 1/ 1     16 x  22 x 512 ->   16 x  22 x 512 0.185 BF
 103 route  102 87                                 ->   16 x  22 x1024
 104 conv   1024       1 x 1/ 1     16 x  22 x1024 ->   16 x  22 x1024 0.738 BF
 105 conv    512       1 x 1/ 1     16 x  22 x1024 ->   16 x  22 x 512 0.369 BF
 106 conv   1024       3 x 3/ 1     16 x  22 x 512 ->   16 x  22 x1024 3.322 BF
 107 conv    512       1 x 1/ 1     16 x  22 x1024 ->   16 x  22 x 512 0.369 BF
 108 max                5x 5/ 1     16 x  22 x 512 ->   16 x  22 x 512 0.005 BF
 109 route  107                                            ->   16 x  22 x 512
 110 max                9x 9/ 1     16 x  22 x 512 ->   16 x  22 x 512 0.015 BF
 111 route  107                                            ->   16 x  22 x 512
 112 max               13x13/ 1     16 x  22 x 512 ->   16 x  22 x 512 0.030 BF
 113 route  112 110 108 107                        ->   16 x  22 x2048
 114 conv    512       1 x 1/ 1     16 x  22 x2048 ->   16 x  22 x 512 0.738 BF
 115 conv   1024       3 x 3/ 1     16 x  22 x 512 ->   16 x  22 x1024 3.322 BF
 116 conv    512       1 x 1/ 1     16 x  22 x1024 ->   16 x  22 x 512 0.369 BF
 117 conv    256       1 x 1/ 1     16 x  22 x 512 ->   16 x  22 x 256 0.092 BF
 118 upsample                 2x    16 x  22 x 256 ->   32 x  44 x 256
 119 route  85                                     ->   32 x  44 x 512
 120 conv    256       1 x 1/ 1     32 x  44 x 512 ->   32 x  44 x 256 0.369 BF
 121 route  120 118                                ->   32 x  44 x 512
 122 conv    256       1 x 1/ 1     32 x  44 x 512 ->   32 x  44 x 256 0.369 BF
 123 conv    512       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 512 3.322 BF
 124 conv    256       1 x 1/ 1     32 x  44 x 512 ->   32 x  44 x 256 0.369 BF
 125 conv    512       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 512 3.322 BF
 126 conv    256       1 x 1/ 1     32 x  44 x 512 ->   32 x  44 x 256 0.369 BF
 127 conv    128       1 x 1/ 1     32 x  44 x 256 ->   32 x  44 x 128 0.092 BF
 128 upsample                 2x    32 x  44 x 128 ->   64 x  88 x 128
 129 route  54                                     ->   64 x  88 x 256
 130 conv    128       1 x 1/ 1     64 x  88 x 256 ->   64 x  88 x 128 0.369 BF
 131 route  130 128                                ->   64 x  88 x 256
 132 conv    128       1 x 1/ 1     64 x  88 x 256 ->   64 x  88 x 128 0.369 BF
 133 conv    256       3 x 3/ 1     64 x  88 x 128 ->   64 x  88 x 256 3.322 BF
 134 conv    128       1 x 1/ 1     64 x  88 x 256 ->   64 x  88 x 128 0.369 BF
 135 conv    256       3 x 3/ 1     64 x  88 x 128 ->   64 x  88 x 256 3.322 BF
 136 conv    128       1 x 1/ 1     64 x  88 x 256 ->   64 x  88 x 128 0.369 BF
 137 conv    256       3 x 3/ 1     64 x  88 x 128 ->   64 x  88 x 256 3.322 BF
 138 conv     18       1 x 1/ 1     64 x  88 x 256 ->   64 x  88 x  18 0.052 BF
 139 yolo
[yolo] params: iou loss: ciou (4), iou_norm: 0.07, cls_norm: 1.00, scale_x_y: 1.20
nms_kind: greedynms (1), beta = 0.600000
 140 route  136                                            ->   64 x  88 x 128
 141 conv    256       3 x 3/ 2     64 x  88 x 128 ->   32 x  44 x 256 0.830 BF
 142 route  141 126                                ->   32 x  44 x 512
 143 conv    256       1 x 1/ 1     32 x  44 x 512 ->   32 x  44 x 256 0.369 BF
 144 conv    512       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 512 3.322 BF
 145 conv    256       1 x 1/ 1     32 x  44 x 512 ->   32 x  44 x 256 0.369 BF
 146 conv    512       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 512 3.322 BF
 147 conv    256       1 x 1/ 1     32 x  44 x 512 ->   32 x  44 x 256 0.369 BF
 148 conv    512       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 512 3.322 BF
 149 conv     18       1 x 1/ 1     32 x  44 x 512 ->   32 x  44 x  18 0.026 BF
 150 yolo
[yolo] params: iou loss: ciou (4), iou_norm: 0.07, cls_norm: 1.00, scale_x_y: 1.10
nms_kind: greedynms (1), beta = 0.600000
 151 route  147                                            ->   32 x  44 x 256
 152 conv    512       3 x 3/ 2     32 x  44 x 256 ->   16 x  22 x 512 0.830 BF
 153 route  152 116                                ->   16 x  22 x1024
 154 conv    512       1 x 1/ 1     16 x  22 x1024 ->   16 x  22 x 512 0.369 BF
 155 conv   1024       3 x 3/ 1     16 x  22 x 512 ->   16 x  22 x1024 3.322 BF
 156 conv    512       1 x 1/ 1     16 x  22 x1024 ->   16 x  22 x 512 0.369 BF
 157 conv   1024       3 x 3/ 1     16 x  22 x 512 ->   16 x  22 x1024 3.322 BF
 158 conv    512       1 x 1/ 1     16 x  22 x1024 ->   16 x  22 x 512 0.369 BF
 159 conv   1024       3 x 3/ 1     16 x  22 x 512 ->   16 x  22 x1024 3.322 BF
 160 conv     18       1 x 1/ 1     16 x  22 x1024 ->   16 x  22 x  18 0.013 BF
 161 yolo
[yolo] params: iou loss: ciou (4), iou_norm: 0.07, cls_norm: 1.00, scale_x_y: 1.05
nms_kind: greedynms (1), beta = 0.600000
Total BFLOPS 124.060
avg_outputs = 1020130
Loading weights from C:\YOLO\YoloV4_Custom_Large.weights...
 seen 64, trained: 640 K-images (10 Kilo-batches_64)
Done! Loaded 162 layers from weights-file
 Used AVX
 Used FMA & AVX2

But if I load up the Tiny model with the _Tiny.cfg and weights files, it finds two bounding boxes:

Tiny_Results

And gives this console output:

Initializing YoloV4_Tiny AI...
If YOLO fails check for Microsoft Visual C++ 2017 Redistributable (x64) vc_redist.x64.exe
batch = 1, time_steps = 1, train = 0
   layer   filters  size/strd(dil)      input                output
   0 conv     32       3 x 3/ 2    512 x 704 x   3 ->  256 x 352 x  32 0.156 BF
   1 conv     64       3 x 3/ 2    256 x 352 x  32 ->  128 x 176 x  64 0.830 BF
   2 conv     64       3 x 3/ 1    128 x 176 x  64 ->  128 x 176 x  64 1.661 BF
   3 route  2                                  1/2 ->  128 x 176 x  32
   4 conv     32       3 x 3/ 1    128 x 176 x  32 ->  128 x 176 x  32 0.415 BF
   5 conv     32       3 x 3/ 1    128 x 176 x  32 ->  128 x 176 x  32 0.415 BF
   6 route  5 4                                    ->  128 x 176 x  64
   7 conv     64       1 x 1/ 1    128 x 176 x  64 ->  128 x 176 x  64 0.185 BF
   8 route  2 7                                    ->  128 x 176 x 128
   9 max                2x 2/ 2    128 x 176 x 128 ->   64 x  88 x 128 0.003 BF
  10 conv    128       3 x 3/ 1     64 x  88 x 128 ->   64 x  88 x 128 1.661 BF
  11 route  10                                 1/2 ->   64 x  88 x  64
  12 conv     64       3 x 3/ 1     64 x  88 x  64 ->   64 x  88 x  64 0.415 BF
  13 conv     64       3 x 3/ 1     64 x  88 x  64 ->   64 x  88 x  64 0.415 BF
  14 route  13 12                                  ->   64 x  88 x 128
  15 conv    128       1 x 1/ 1     64 x  88 x 128 ->   64 x  88 x 128 0.185 BF
  16 route  10 15                                  ->   64 x  88 x 256
  17 max                2x 2/ 2     64 x  88 x 256 ->   32 x  44 x 256 0.001 BF
  18 conv    256       3 x 3/ 1     32 x  44 x 256 ->   32 x  44 x 256 1.661 BF
  19 route  18                                 1/2 ->   32 x  44 x 128
  20 conv    128       3 x 3/ 1     32 x  44 x 128 ->   32 x  44 x 128 0.415 BF
  21 conv    128       3 x 3/ 1     32 x  44 x 128 ->   32 x  44 x 128 0.415 BF
  22 route  21 20                                  ->   32 x  44 x 256
  23 conv    256       1 x 1/ 1     32 x  44 x 256 ->   32 x  44 x 256 0.185 BF
  24 route  18 23                                  ->   32 x  44 x 512
  25 max                2x 2/ 2     32 x  44 x 512 ->   16 x  22 x 512 0.001 BF
  26 conv    512       3 x 3/ 1     16 x  22 x 512 ->   16 x  22 x 512 1.661 BF
  27 conv    256       1 x 1/ 1     16 x  22 x 512 ->   16 x  22 x 256 0.092 BF
  28 conv    512       3 x 3/ 1     16 x  22 x 256 ->   16 x  22 x 512 0.830 BF
  29 conv     18       1 x 1/ 1     16 x  22 x 512 ->   16 x  22 x  18 0.006 BF
  30 yolo
[yolo] params: iou loss: ciou (4), iou_norm: 0.07, cls_norm: 1.00, scale_x_y: 1.05
nms_kind: greedynms (1), beta = 0.600000
Unused field: 'resize = 1.5'
  31 route  27                                     ->   16 x  22 x 256
  32 conv    128       1 x 1/ 1     16 x  22 x 256 ->   16 x  22 x 128 0.023 BF
  33 upsample                 2x    16 x  22 x 128 ->   32 x  44 x 128
  34 route  33 23                                  ->   32 x  44 x 384
  35 conv    256       3 x 3/ 1     32 x  44 x 384 ->   32 x  44 x 256 2.491 BF
  36 conv     18       1 x 1/ 1     32 x  44 x 256 ->   32 x  44 x  18 0.013 BF
  37 yolo
[yolo] params: iou loss: ciou (4), iou_norm: 0.07, cls_norm: 1.00, scale_x_y: 1.05
nms_kind: greedynms (1), beta = 0.600000
Unused field: 'resize = 1.5'
Total BFLOPS 14.137
avg_outputs = 624151
Loading weights from C:\YOLO\YoloV4_Custom_Tiny_3_22_2022.weights...
 seen 64, trained: 3840 K-images (60 Kilo-batches_64)
Done! Loaded 38 layers from weights-file
 Used AVX
 Used FMA & AVX2
Result [ processed in 634 ms ]

My method for initializing the Yolo weights in C#:

private void InitializeYoloModel(YoloConfiguration config)
        {
            try
            {
                if (this.YoloModel!= null)
                {
                    this.YoloModel.Dispose();
                }

                var useOnlyCpu = true;
                Console.WriteLine("If YOLO fails check for Microsoft Visual C++ 2017 Redistributable (x64) vc_redist.x64.exe");
                var sw = new Stopwatch();
                sw.Start();
                this.YoloModel = new YoloWrapper(config); // , 0, useOnlyCPU
                sw.Stop();

                var action = new MethodInvoker(delegate ()
                {
                    var detectionSystemDetail = string.Empty;
                    if (!string.IsNullOrEmpty(DetectionSystem.GPU.ToString()))
                    {
                        detectionSystemDetail = $"({DetectionSystem.GPU.ToString()})";
                    }
                    Console.WriteLine($"Initialize YoloModel in {sw.Elapsed.TotalMilliseconds:0} ms - Detection System:{this.YoloModel.DetectionSystem} {detectionSystemDetail} Weights:{config.WeightsFile}");
                });
            }
            catch (Exception ee)
            {
                Console.WriteLine("Error loading YoloModel" + ee.Message);
                YoloModel = null;
            }
        }

As an extra note, I don't have any Microsoft Visual C++ Redistributable versions beyond this one:

Microsoft Visual C++ 2015-2019 Redistributable (x64) - 14.22.27821

As I said, my configuration allows me to get the program working without issue with the Tiny model, the large model even initializes without returning any errors, it just doesn't return predictions as I would expect based on its performance in darknet.exe.

If there is any additional information that would be helpful to determine the cause of this failure, please let me know and I will provide it for you asap.

Thanks so much for your time.

EDIT: I have also loaded the provided YoloV3 pretrained model and have gotten the predictions I expect from that pretrained model, still not sure where the point of failure is for my custom-trained large YoloV4 model.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@XononoX