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

Support for YOLO v5 / v6 / v7 or other variants #26

Closed
caitou2019 opened this issue Mar 31, 2022 · 5 comments
Closed

Support for YOLO v5 / v6 / v7 or other variants #26

caitou2019 opened this issue Mar 31, 2022 · 5 comments
Assignees
@LdDl
Labels
discussion enhancement New feature or request

Comments

@caitou2019
Copy link

caitou2019 commented Mar 31, 2022 •
edited
Loading

Do we have any plans to support Yolo-V5 in this project?
@caitou2019 caitou2019 added the enhancement New feature or request label Mar 31, 2022
@caitou2019 caitou2019 changed the title [FEATURE REQUEST]support yoloV5 support yoloV5 Mar 31, 2022
@LdDl
Copy link
Owner

LdDl commented Mar 31, 2022 •
edited
Loading

Isn't YOLOv5 written in Python+Pytroch natively? I guess so...
C-code must be happen to create proper CGO bindings :)

There are some interesting readings:

  1. Repo Claims To Be YOLOv5 AlexeyAB/darknet#5920 (comment)
  2. https://blog.roboflow.com/yolov4-versus-yolov5/

I don't think v5 will be supported by actual maintainers of Darknet

@LdDl LdDl added the discussion label Apr 2, 2022
@LdDl
Copy link
Owner

LdDl commented Jul 7, 2022 •
edited
Loading

Well
Now I have plans for YoloV7 since this is official - AlexeyAB/darknet#8595

@LdDl LdDl changed the title support yoloV5 support yoloV5 / v6 / v7? Jul 7, 2022
@LdDl LdDl changed the title support yoloV5 / v6 / v7? Support for YOLO v5 / v6 / v7 or other variants Jul 7, 2022
@TaihuLight
Copy link

@LdDl @AlexeyAB It is a very good new that the darknet will support the training and inference of YOLOV7.

@LdDl
Copy link
Owner

LdDl commented Jul 13, 2022

YOLOv7 tiny should work without additional effort: 1969bda

~/go_work/go-darknet/cmd/examples$ go build -o base_example/main base_example/main.go && ./base_example/main --configFile=yolov7-tiny.cfg --weightsFile=yolov7-tiny.weights --imageFile=sample.jpg
 Try to load cfg: yolov7-tiny.cfg, clear = 0 
 0 : compute_capability = 860, cudnn_half = 0, GPU: NVIDIA GeForce RTX 3060 
net.optimized_memory = 0 
mini_batch = 1, batch = 1, time_steps = 1, train = 1 
   layer   filters  size/strd(dil)      input                output
   0 Create CUDA-stream - 0 
 Create cudnn-handle 0 
conv     32       3 x 3/ 2    416 x 416 x   3 ->  208 x 208 x  32 0.075 BF
   1 conv     64       3 x 3/ 2    208 x 208 x  32 ->  104 x 104 x  64 0.399 BF
   2 conv     32       1 x 1/ 1    104 x 104 x  64 ->  104 x 104 x  32 0.044 BF
   3 route  1                                      ->  104 x 104 x  64 
   4 conv     32       1 x 1/ 1    104 x 104 x  64 ->  104 x 104 x  32 0.044 BF
   5 conv     32       3 x 3/ 1    104 x 104 x  32 ->  104 x 104 x  32 0.199 BF
   6 conv     32       3 x 3/ 1    104 x 104 x  32 ->  104 x 104 x  32 0.199 BF
   7 route  2 4 5 6                        ->  104 x 104 x 128 
   8 conv     64       1 x 1/ 1    104 x 104 x 128 ->  104 x 104 x  64 0.177 BF
   9 max                2x 2/ 2    104 x 104 x  64 ->   52 x  52 x  64 0.001 BF
  10 conv     64       1 x 1/ 1     52 x  52 x  64 ->   52 x  52 x  64 0.022 BF
  11 route  9                                      ->   52 x  52 x  64 
  12 conv     64       1 x 1/ 1     52 x  52 x  64 ->   52 x  52 x  64 0.022 BF
  13 conv     64       3 x 3/ 1     52 x  52 x  64 ->   52 x  52 x  64 0.199 BF
  14 conv     64       3 x 3/ 1     52 x  52 x  64 ->   52 x  52 x  64 0.199 BF
  15 route  10 12 13 14                            ->   52 x  52 x 256 
  16 conv    128       1 x 1/ 1     52 x  52 x 256 ->   52 x  52 x 128 0.177 BF
  17 max                2x 2/ 2     52 x  52 x 128 ->   26 x  26 x 128 0.000 BF
  18 conv    128       1 x 1/ 1     26 x  26 x 128 ->   26 x  26 x 128 0.022 BF
  19 route  17                                     ->   26 x  26 x 128 
  20 conv    128       1 x 1/ 1     26 x  26 x 128 ->   26 x  26 x 128 0.022 BF
  21 conv    128       3 x 3/ 1     26 x  26 x 128 ->   26 x  26 x 128 0.199 BF
  22 conv    128       3 x 3/ 1     26 x  26 x 128 ->   26 x  26 x 128 0.199 BF
  23 route  18 20 21 22                            ->   26 x  26 x 512 
  24 conv    256       1 x 1/ 1     26 x  26 x 512 ->   26 x  26 x 256 0.177 BF
  25 max                2x 2/ 2     26 x  26 x 256 ->   13 x  13 x 256 0.000 BF
  26 conv    256       1 x 1/ 1     13 x  13 x 256 ->   13 x  13 x 256 0.022 BF
  27 route  25                                     ->   13 x  13 x 256 
  28 conv    256       1 x 1/ 1     13 x  13 x 256 ->   13 x  13 x 256 0.022 BF
  29 conv    256       3 x 3/ 1     13 x  13 x 256 ->   13 x  13 x 256 0.199 BF
  30 conv    256       3 x 3/ 1     13 x  13 x 256 ->   13 x  13 x 256 0.199 BF
  31 route  26 28 29 30                            ->   13 x  13 x1024 
  32 conv    512       1 x 1/ 1     13 x  13 x1024 ->   13 x  13 x 512 0.177 BF
  33 conv    256       1 x 1/ 1     13 x  13 x 512 ->   13 x  13 x 256 0.044 BF
  34 route  32                                     ->   13 x  13 x 512 
  35 conv    256       1 x 1/ 1     13 x  13 x 512 ->   13 x  13 x 256 0.044 BF
  36 max                5x 5/ 1     13 x  13 x 256 ->   13 x  13 x 256 0.001 BF
  37 route  35                                     ->   13 x  13 x 256 
  38 max                9x 9/ 1     13 x  13 x 256 ->   13 x  13 x 256 0.004 BF
  39 route  35                                     ->   13 x  13 x 256 
  40 max               13x13/ 1     13 x  13 x 256 ->   13 x  13 x 256 0.007 BF
  41 route  40 38 36 35                            ->   13 x  13 x1024 
  42 conv    256       1 x 1/ 1     13 x  13 x1024 ->   13 x  13 x 256 0.089 BF
  43 route  33 42                                  ->   13 x  13 x 512 
  44 conv    256       1 x 1/ 1     13 x  13 x 512 ->   13 x  13 x 256 0.044 BF
  45 conv    128       1 x 1/ 1     13 x  13 x 256 ->   13 x  13 x 128 0.011 BF
  46 upsample                 2x    13 x  13 x 128 ->   26 x  26 x 128
  47 route  24                                     ->   26 x  26 x 256 
  48 conv    128       1 x 1/ 1     26 x  26 x 256 ->   26 x  26 x 128 0.044 BF
  49 route  48 46                                  ->   26 x  26 x 256 
  50 conv     64       1 x 1/ 1     26 x  26 x 256 ->   26 x  26 x  64 0.022 BF
  51 route  49                                     ->   26 x  26 x 256 
  52 conv     64       1 x 1/ 1     26 x  26 x 256 ->   26 x  26 x  64 0.022 BF
  53 conv     64       3 x 3/ 1     26 x  26 x  64 ->   26 x  26 x  64 0.050 BF
  54 conv     64       3 x 3/ 1     26 x  26 x  64 ->   26 x  26 x  64 0.050 BF
  55 route  50 52 53 54                            ->   26 x  26 x 256 
  56 conv    128       1 x 1/ 1     26 x  26 x 256 ->   26 x  26 x 128 0.044 BF
  57 conv     64       1 x 1/ 1     26 x  26 x 128 ->   26 x  26 x  64 0.011 BF
  58 upsample                 2x    26 x  26 x  64 ->   52 x  52 x  64
  59 route  16                                     ->   52 x  52 x 128 
  60 conv     64       1 x 1/ 1     52 x  52 x 128 ->   52 x  52 x  64 0.044 BF
  61 route  60 58                                  ->   52 x  52 x 128 
  62 conv     32       1 x 1/ 1     52 x  52 x 128 ->   52 x  52 x  32 0.022 BF
  63 route  61                                     ->   52 x  52 x 128 
  64 conv     32       1 x 1/ 1     52 x  52 x 128 ->   52 x  52 x  32 0.022 BF
  65 conv     32       3 x 3/ 1     52 x  52 x  32 ->   52 x  52 x  32 0.050 BF
  66 conv     32       3 x 3/ 1     52 x  52 x  32 ->   52 x  52 x  32 0.050 BF
  67 route  62 64 65 66                            ->   52 x  52 x 128 
  68 conv     64       1 x 1/ 1     52 x  52 x 128 ->   52 x  52 x  64 0.044 BF
  69 conv    128       3 x 3/ 2     52 x  52 x  64 ->   26 x  26 x 128 0.100 BF
  70 route  69 56                                  ->   26 x  26 x 256 
  71 conv     64       1 x 1/ 1     26 x  26 x 256 ->   26 x  26 x  64 0.022 BF
  72 route  70                                     ->   26 x  26 x 256 
  73 conv     64       1 x 1/ 1     26 x  26 x 256 ->   26 x  26 x  64 0.022 BF
  74 conv     64       3 x 3/ 1     26 x  26 x  64 ->   26 x  26 x  64 0.050 BF
  75 conv     64       3 x 3/ 1     26 x  26 x  64 ->   26 x  26 x  64 0.050 BF
  76 route  71 73 74 75                            ->   26 x  26 x 256 
  77 conv    128       1 x 1/ 1     26 x  26 x 256 ->   26 x  26 x 128 0.044 BF
  78 conv    256       3 x 3/ 2     26 x  26 x 128 ->   13 x  13 x 256 0.100 BF
  79 route  78 44                                  ->   13 x  13 x 512 
  80 conv    128       1 x 1/ 1     13 x  13 x 512 ->   13 x  13 x 128 0.022 BF
  81 route  79                                     ->   13 x  13 x 512 
  82 conv    128       1 x 1/ 1     13 x  13 x 512 ->   13 x  13 x 128 0.022 BF
  83 conv    128       3 x 3/ 1     13 x  13 x 128 ->   13 x  13 x 128 0.050 BF
  84 conv    128       3 x 3/ 1     13 x  13 x 128 ->   13 x  13 x 128 0.050 BF
  85 route  80 82 83 84                            ->   13 x  13 x 512 
  86 conv    256       1 x 1/ 1     13 x  13 x 512 ->   13 x  13 x 256 0.044 BF
  87 route  68                                     ->   52 x  52 x  64 
  88 conv    128       3 x 3/ 1     52 x  52 x  64 ->   52 x  52 x 128 0.399 BF
  89 conv    255       1 x 1/ 1     52 x  52 x 128 ->   52 x  52 x 255 0.177 BF
  90 yolo
[yolo] params: iou loss: ciou (4), iou_norm: 0.05, obj_norm: 1.00, cls_norm: 0.50, delta_norm: 1.00, scale_x_y: 2.00
nms_kind: diounms (2), beta = 0.600000 
  91 route  77                                     ->   26 x  26 x 128 
  92 conv    256       3 x 3/ 1     26 x  26 x 128 ->   26 x  26 x 256 0.399 BF
  93 conv    255       1 x 1/ 1     26 x  26 x 256 ->   26 x  26 x 255 0.088 BF
  94 yolo
[yolo] params: iou loss: ciou (4), iou_norm: 0.05, obj_norm: 1.00, cls_norm: 0.50, delta_norm: 1.00, scale_x_y: 2.00
nms_kind: diounms (2), beta = 0.600000 
  95 route  86                                     ->   13 x  13 x 256 
  96 conv    512       3 x 3/ 1     13 x  13 x 256 ->   13 x  13 x 512 0.399 BF
  97 conv    255       1 x 1/ 1     13 x  13 x 512 ->   13 x  13 x 255 0.044 BF
  98 yolo
[yolo] params: iou loss: ciou (4), iou_norm: 0.05, obj_norm: 1.00, cls_norm: 0.50, delta_norm: 1.00, scale_x_y: 2.00
nms_kind: diounms (2), beta = 0.600000 
Unused field: 'names = coco.names'
Total BFLOPS 5.802 
avg_outputs = 182380 
 Allocate additional workspace_size = 9.34 MB 
 Try to load weights: yolov7-tiny.weights 
Loading weights from yolov7-tiny.weights...
 seen 64, trained: 0 K-images (0 Kilo-batches_64) 
Done! Loaded 99 layers from weights-file 
Loaded - names_list: coco.names, classes = 80 
2022/07/13 16:13:15 Network-only time taken: 7.88269ms
2022/07/13 16:13:15 Overall time taken: 8.386104ms 101
truck (7): 53.2890% | start point: (0,143) | end point: (89, 328)
truck (7): 42.1364% | start point: (685,182) | end point: (800, 318)
truck (7): 26.9703% | start point: (437,170) | end point: (560, 217)
car (2): 87.7818% | start point: (509,189) | end point: (742, 329)
car (2): 87.5633% | start point: (262,191) | end point: (423, 322)
car (2): 85.4743% | start point: (427,198) | end point: (549, 309)
car (2): 71.3772% | start point: (0,147) | end point: (87, 327)
car (2): 62.5698% | start point: (98,151) | end point: (197, 286)
car (2): 61.5811% | start point: (693,186) | end point: (799, 316)
car (2): 49.6343% | start point: (386,206) | end point: (441, 286)
car (2): 28.2012% | start point: (386,205) | end point: (440, 236)
bicycle (1): 71.9609% | start point: (179,294) | end point: (249, 405)
person (0): 85.4390% | start point: (146,130) | end point: (269, 351)

@LdDl
Copy link
Owner

LdDl commented Jul 13, 2022

Comparing to YOLO v4 tiny:

~/go_work/go-darknet/cmd/examples$ go build -o base_example/main base_example/main.go && ./base_example/main --configFile=yolov4-tiny.cfg --weightsFile=yolov4-tiny.weights --imageFile=sample.jpg
 Try to load cfg: yolov4-tiny.cfg, clear = 0 
 0 : compute_capability = 860, cudnn_half = 0, GPU: NVIDIA GeForce RTX 3060 
net.optimized_memory = 0 
mini_batch = 1, batch = 1, time_steps = 1, train = 1 
   layer   filters  size/strd(dil)      input                output
   0 Create CUDA-stream - 0 
 Create cudnn-handle 0 
conv     32       3 x 3/ 2    416 x 416 x   3 ->  208 x 208 x  32 0.075 BF
   1 conv     64       3 x 3/ 2    208 x 208 x  32 ->  104 x 104 x  64 0.399 BF
   2 conv     64       3 x 3/ 1    104 x 104 x  64 ->  104 x 104 x  64 0.797 BF
   3 route  2                                  1/2 ->  104 x 104 x  32 
   4 conv     32       3 x 3/ 1    104 x 104 x  32 ->  104 x 104 x  32 0.199 BF
   5 conv     32       3 x 3/ 1    104 x 104 x  32 ->  104 x 104 x  32 0.199 BF
   6 route  5 4                                    ->  104 x 104 x  64 
   7 conv     64       1 x 1/ 1    104 x 104 x  64 ->  104 x 104 x  64 0.089 BF
   8 route  2 7                                    ->  104 x 104 x 128 
   9 max                2x 2/ 2    104 x 104 x 128 ->   52 x  52 x 128 0.001 BF
  10 conv    128       3 x 3/ 1     52 x  52 x 128 ->   52 x  52 x 128 0.797 BF
  11 route  10                                 1/2 ->   52 x  52 x  64 
  12 conv     64       3 x 3/ 1     52 x  52 x  64 ->   52 x  52 x  64 0.199 BF
  13 conv     64       3 x 3/ 1     52 x  52 x  64 ->   52 x  52 x  64 0.199 BF
  14 route  13 12                                  ->   52 x  52 x 128 
  15 conv    128       1 x 1/ 1     52 x  52 x 128 ->   52 x  52 x 128 0.089 BF
  16 route  10 15                                  ->   52 x  52 x 256 
  17 max                2x 2/ 2     52 x  52 x 256 ->   26 x  26 x 256 0.001 BF
  18 conv    256       3 x 3/ 1     26 x  26 x 256 ->   26 x  26 x 256 0.797 BF
  19 route  18                                 1/2 ->   26 x  26 x 128 
  20 conv    128       3 x 3/ 1     26 x  26 x 128 ->   26 x  26 x 128 0.199 BF
  21 conv    128       3 x 3/ 1     26 x  26 x 128 ->   26 x  26 x 128 0.199 BF
  22 route  21 20                                  ->   26 x  26 x 256 
  23 conv    256       1 x 1/ 1     26 x  26 x 256 ->   26 x  26 x 256 0.089 BF
  24 route  18 23                                  ->   26 x  26 x 512 
  25 max                2x 2/ 2     26 x  26 x 512 ->   13 x  13 x 512 0.000 BF
  26 conv    512       3 x 3/ 1     13 x  13 x 512 ->   13 x  13 x 512 0.797 BF
  27 conv    256       1 x 1/ 1     13 x  13 x 512 ->   13 x  13 x 256 0.044 BF
  28 conv    512       3 x 3/ 1     13 x  13 x 256 ->   13 x  13 x 512 0.399 BF
  29 conv    255       1 x 1/ 1     13 x  13 x 512 ->   13 x  13 x 255 0.044 BF
  30 yolo
[yolo] params: iou loss: ciou (4), iou_norm: 0.07, obj_norm: 1.00, cls_norm: 1.00, delta_norm: 1.00, scale_x_y: 1.05
nms_kind: greedynms (1), beta = 0.600000 
  31 route  27                                     ->   13 x  13 x 256 
  32 conv    128       1 x 1/ 1     13 x  13 x 256 ->   13 x  13 x 128 0.011 BF
  33 upsample                 2x    13 x  13 x 128 ->   26 x  26 x 128
  34 route  33 23                                  ->   26 x  26 x 384 
  35 conv    256       3 x 3/ 1     26 x  26 x 384 ->   26 x  26 x 256 1.196 BF
  36 conv    255       1 x 1/ 1     26 x  26 x 256 ->   26 x  26 x 255 0.088 BF
  37 yolo
[yolo] params: iou loss: ciou (4), iou_norm: 0.07, obj_norm: 1.00, cls_norm: 1.00, delta_norm: 1.00, scale_x_y: 1.05
nms_kind: greedynms (1), beta = 0.600000 
Unused field: 'names = coco.names'
Total BFLOPS 6.910 
avg_outputs = 310203 
 Allocate additional workspace_size = 13.80 MB 
 Try to load weights: yolov4-tiny.weights 
Loading weights from yolov4-tiny.weights...
 seen 64, trained: 0 K-images (0 Kilo-batches_64) 
Done! Loaded 38 layers from weights-file 
Loaded - names_list: coco.names, classes = 80 
2022/07/13 16:19:28 Network-only time taken: 6.694944ms
2022/07/13 16:19:28 Overall time taken: 6.854318ms 30
truck (7): 77.7936% | start point: (0,138) | end point: (90, 332)
truck (7): 55.9773% | start point: (696,174) | end point: (799, 314)
car (2): 53.1286% | start point: (696,184) | end point: (799, 319)
car (2): 98.0222% | start point: (262,189) | end point: (424, 330)
car (2): 97.8773% | start point: (430,190) | end point: (542, 313)
car (2): 81.4099% | start point: (510,190) | end point: (743, 325)
car (2): 43.3935% | start point: (391,207) | end point: (435, 299)
car (2): 37.4221% | start point: (386,206) | end point: (429, 239)
car (2): 32.0724% | start point: (109,196) | end point: (157, 289)
person (0): 73.0868% | start point: (154,132) | end point: (284, 382)

I see that accuracy improvements in v7-tiny variation, but inference a bit fast in v4-tiny

@LdDl LdDl closed this as completed Jun 26, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@LdDl LdDl

Labels
discussion enhancement New feature or request
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
3 participants
@LdDl @TaihuLight @caitou2019