darknet_AlexeyAB_191221 - bag
1. anchors
2. p*****_yolov3_bag.data
/home/yongqiang/darknet_work/darknet_AlexeyAB_191221/darknet-master/data/p*****_yolov3_bag.data
classes= 3
train = /media/famu/DISK_DEEP/yongqiang/p*****_yolov3_bag/training_validation_test_sets_bag_v1.txt
# valid = /media/famu/DISK_DEEP/yongqiang/p*****_yolov3_bag/training_validation_test_sets_bag_v0.txt
valid = /media/famu/DISK_DEEP/yongqiang/p*****_yolov3_bag/training_validation_test_sets_bag_v0.txt
names = data/p*****_yolov3_bag.names
backup = /media/famu/DISK_DATA/yongqiang/backup_p*****_yolov3_bag
eval=coco
3. p*****_yolov3_bag.names
/home/yongqiang/darknet_work/darknet_AlexeyAB_191221/darknet-master/data/p*****_yolov3_bag.names
backpack
handbag
suitcase
4. To calculate anchors
./darknet detector calc_anchors data/p*****_yolov3_bag.data -num_of_clusters 8 -width 608 -height 608
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$ ./darknet detector calc_anchors data/p*****_yolov3_bag.data -num_of_clusters 8 -width 608 -height 608
num_of_clusters = 8, width = 608, height = 608
read labels from 12001 images
loaded image: 12001 box: 38262
all loaded.
calculating k-means++ ...
iterations = 88
avg IoU = 69.22 %
Saving anchors to the file: anchors.txt
anchors = 13, 30, 26, 54, 47, 85, 98, 89, 86,190, 187,147, 238,315, 456,494
^C
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$
5. Training - ./darknet detector train ./train_cfg/p*****_yolov3_bag.data ./train_cfg/p*****_yolov3_bag.cfg -gpus 0,1,2,3 -map
Just train with -map flag:
darknet.exe detector train data/obj.data yolo-obj.cfg darknet53.conv.74 -map
So you will see mAP-chart (red-line) in the Loss-chart Window. mAP will be calculated for each 4 Epochs using valid=valid.txt file that is specified in obj.data file (1 Epoch = images_in_train_txt / batch iterations)
(to change the max x-axis value - change max_batches= parameter to 2000*classes, f.e. max_batches=6000 for 3 classes)
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$ ./darknet detector train ./train_cfg/p*****_yolov3_bag.data ./train_cfg/p*****_yolov3_bag.cfg -gpus 0,1,2,3 -map
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 175 Avg (IOU: 0.921591, GIOU: 0.920626), Class: 0.999561, Obj: 0.840744, No Obj: 0.000146, .5R: 1.000000, .75R: 1.000000, count: 3, loss = 0.248484, class_loss = 0.223862, iou_loss = 0.024622
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 167 Avg (IOU: 0.863015, GIOU: 0.861838), Class: 0.995141, Obj: 0.484292, No Obj: 0.001002, .5R: 1.000000, .75R: 1.000000, count: 4, loss = 0.918329, class_loss = 0.488470, iou_loss = 0.429858
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 175 Avg (IOU: -nan, GIOU: -nan), Class: -nan, Obj: -nan, No Obj: 0.000006, .5R: -nan, .75R: -nan, count: 0, loss = 0.005137, class_loss = 0.005137, iou_loss = 0.000000
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 167 Avg (IOU: -nan, GIOU: -nan), Class: -nan, Obj: -nan, No Obj: 0.000000, .5R: -nan, .75R: -nan, count: 0, loss = 0.000000, class_loss = 0.000000, iou_loss = 0.000000
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 175 Avg (IOU: 0.817482, GIOU: 0.814469), Class: 0.983876, Obj: 0.542657, No Obj: 0.000341, .5R: 1.000000, .75R: 0.916667, count: 12, loss = 2.177800, class_loss = 1.450139, iou_loss = 0.727660
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 167 Avg (IOU: -nan, GIOU: -nan), Class: -nan, Obj: -nan, No Obj: 0.000001, .5R: -nan, .75R: -nan, count: 0, loss = 0.000000, class_loss = 0.000000, iou_loss = 0.000000
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 167 Avg (IOU: -nan, GIOU: -nan), Class: -nan, Obj: -nan, No Obj: 0.000021, .5R: -nan, .75R: -nan, count: 0, loss = 0.001184, class_loss = 0.001184, iou_loss = 0.000000
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 175 Avg (IOU: 0.730559, GIOU: 0.721980), Class: 0.943102, Obj: 0.500754, No Obj: 0.000256, .5R: 1.000000, .75R: 0.428571, count: 14, loss = 2.817832, class_loss = 1.911624, iou_loss = 0.906208
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 167 Avg (IOU: -nan, GIOU: -nan), Class: -nan, Obj: -nan, No Obj: 0.000033, .5R: -nan, .75R: -nan, count: 0, loss = 0.000002, class_loss = 0.000002, iou_loss = 0.000000
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 175 Avg (IOU: 0.738928, GIOU: 0.727626), Class: 0.907107, Obj: 0.594643, No Obj: 0.000725, .5R: 0.945946, .75R: 0.540541, count: 37, loss = 6.891733, class_loss = 4.579192, iou_loss = 2.312541
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 167 Avg (IOU: 0.889246, GIOU: 0.888135), Class: 0.997077, Obj: 0.034404, No Obj: 0.000069, .5R: 1.000000, .75R: 1.000000, count: 2, loss = 0.625570, class_loss = 0.492146, iou_loss = 0.133424
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 175 Avg (IOU: 0.803413, GIOU: 0.800832), Class: 0.994416, Obj: 0.617647, No Obj: 0.000494, .5R: 0.933333, .75R: 0.800000, count: 15, loss = 2.334442, class_loss = 1.446612, iou_loss = 0.887830
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 175 Avg (IOU: 0.761393, GIOU: 0.749543), Class: 0.909525, Obj: 0.544146, No Obj: 0.000294, .5R: 1.000000, .75R: 0.545455, count: 11, loss = 2.501634, class_loss = 1.852066, iou_loss = 0.649567
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 167 Avg (IOU: 0.852430, GIOU: 0.846127), Class: 0.999595, Obj: 0.996625, No Obj: 0.000702, .5R: 1.000000, .75R: 1.000000, count: 1, loss = 0.019382, class_loss = 0.000006, iou_loss = 0.019375
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 175 Avg (IOU: 0.758557, GIOU: 0.753923), Class: 0.697662, Obj: 0.313631, No Obj: 0.000072, .5R: 1.000000, .75R: 0.333333, count: 3, loss = 1.205699, class_loss = 0.944062, iou_loss = 0.261637
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 167 Avg (IOU: 0.891751, GIOU: 0.891707), Class: 0.994823, Obj: 0.638932, No Obj: 0.000654, .5R: 1.000000, .75R: 1.000000, count: 2, loss = 0.434503, class_loss = 0.342994, iou_loss = 0.091509
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 175 Avg (IOU: 0.676212, GIOU: 0.657095), Class: 0.977033, Obj: 0.290804, No Obj: 0.000118, .5R: 1.000000, .75R: 0.166667, count: 6, loss = 1.957357, class_loss = 1.155171, iou_loss = 0.802186
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 167 Avg (IOU: 0.900844, GIOU: 0.898482), Class: 0.999967, Obj: 0.946276, No Obj: 0.000780, .5R: 1.000000, .75R: 1.000000, count: 2, loss = 0.035735, class_loss = 0.003005, iou_loss = 0.032730
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 175 Avg (IOU: 0.713699, GIOU: 0.703042), Class: 0.856487, Obj: 0.496504, No Obj: 0.000713, .5R: 0.935484, .75R: 0.419355, count: 31, loss = 7.429474, class_loss = 4.887430, iou_loss = 2.542044
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 167 Avg (IOU: -nan, GIOU: -nan), Class: -nan, Obj: -nan, No Obj: 0.000001, .5R: -nan, .75R: -nan, count: 0, loss = 0.000005, class_loss = 0.000005, iou_loss = 0.000000
v3 (mse loss, Normalizer: (iou: 0.750000, cls: 1.000000) Region 175 Avg (IOU: 0.781718, GIOU: 0.774675), Class: 0.956640, Obj: 0.628695, No Obj: 0.000517, .5R: 0.950000, .75R: 0.750000, count: 20, loss = 2.988275, class_loss = 1.863192, iou_loss = 1.125083
Syncing... Done!
(next mAP calculation at 210814 iterations)
Last accuracy mAP@0.5 = 64.07 %, best = 65.70 %
210208: 1.086927, 0.992169 avg loss, 0.000040 rate, 1.937609 seconds, 53813248 images
Saving weights to /media/famu/DISK_DATA/yongqiang/backup_p*****_yolov3_bag/p*****_yolov3_bag_final.weights
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$ ./darknet detector train ./train_cfg/p*****_yolov3_bag.data ./train_cfg/p*****_yolov3_bag.cfg -gpus 0,1,2,3 -map
6. mAP
[net]
# Testing
# batch=1
# subdivisions=1
# Training
batch=64
subdivisions=16
width=608
height=608
channels=3
momentum=0.9
decay=0.0005
angle=0
saturation = 1.5
exposure = 1.5
hue=.1
learning_rate=0.001
burn_in=1000
max_batches = 210200
policy=steps
steps=90000,160000
scales=.1,.1
First - 002
network resolution = 608 * 608 mean average precision (mAP@0.50) = 0.656973, or 65.70 %
network resolution = 544 * 544 mean average precision (mAP@0.50) = 0.643579, or 64.36 %
network resolution = 480 * 480 mean average precision (mAP@0.50) = 0.631074, or 63.11 %
network resolution = 416 * 416 mean average precision (mAP@0.50) = 0.602397, or 60.24 %
network resolution = 352 * 352 mean average precision (mAP@0.50) = 0.545619, or 54.56 %
Second - 003
network resolution = 608 * 608 mean average precision (mAP@0.50) = 0.653740, or 65.37 %
network resolution = 544 * 544 mean average precision (mAP@0.50) = 0.637983, or 63.80 %
network resolution = 480 * 480 mean average precision (mAP@0.50) = 0.625842, or 62.58 %
network resolution = 416 * 416 mean average precision (mAP@0.50) = 0.598726, or 59.87 %
network resolution = 352 * 352 mean average precision (mAP@0.50) = 0.541292, or 54.13 %
6.1 network resolution = 608 * 608
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$ ./darknet detector map ./cfg/p*****_yolov3_bag.data ./cfg/p*****_yolov3_bag.cfg /media/famu/DISK_DATA/yongqiang/p*****_yolov3_bag_best.weights -i 3
net.optimized_memory = 0
batch = 1, time_steps = 1, train = 0
layer filters size/strd(dil) input output
0 conv 32 3 x 3/ 2 608 x 608 x 3 -> 304 x 304 x 32 0.160 BF
1 conv 16 1 x 1/ 1 304 x 304 x 32 -> 304 x 304 x 16 0.095 BF
2 conv 32 3 x 3/ 2 304 x 304 x 16 -> 152 x 152 x 32 0.213 BF
......
Total BFLOPS 15.074
Allocate additional workspace_size = 13.11 MB
Loading weights from /media/famu/DISK_DATA/yongqiang/p*****_yolov3_bag_best.weights...
seen 64
Done! Loaded 176 layers from weights-file
calculation mAP (mean average precision)...
1204
detections_count = 8448, unique_truth_count = 2997
class_id = 0, name = backpack, ap = 76.51% (TP = 774, FP = 165)
class_id = 1, name = handbag, ap = 45.73% (TP = 457, FP = 237)
class_id = 2, name = suitcase, ap = 74.85% (TP = 491, FP = 115)
for conf_thresh = 0.25, precision = 0.77, recall = 0.57, F1-score = 0.66
for conf_thresh = 0.25, TP = 1722, FP = 517, FN = 1275, average IoU = 61.13 %
IoU threshold = 50 %, used Area-Under-Curve for each unique Recall
mean average precision (mAP@0.50) = 0.656973, or 65.70 %
Total Detection Time: 17.000000 Seconds
Set -points flag:
`-points 101` for MS COCO
`-points 11` for PascalVOC 2007 (uncomment `difficult` in voc.data)
`-points 0` (AUC) for ImageNet, PascalVOC 2010-2012, your custom dataset
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$
6.2 network resolution = 544 * 544
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$ ./darknet detector map ./cfg/p*****_yolov3_bag.data ./cfg/p*****_yolov3_bag.cfg /media/famu/DISK_DATA/yongqiang/p*****_yolov3_bag_best.weights -i 3
net.optimized_memory = 0
batch = 1, time_steps = 1, train = 0
layer filters size/strd(dil) input output
0 conv 32 3 x 3/ 2 544 x 544 x 3 -> 272 x 272 x 32 0.128 BF
1 conv 16 1 x 1/ 1 272 x 272 x 32 -> 272 x 272 x 16 0.076 BF
2 conv 32 3 x 3/ 2 272 x 272 x 16 -> 136 x 136 x 32 0.170 BF
......
Total BFLOPS 12.067
Allocate additional workspace_size = 13.11 MB
Loading weights from /media/famu/DISK_DATA/yongqiang/p*****_yolov3_bag_best.weights...
seen 64
Done! Loaded 176 layers from weights-file
calculation mAP (mean average precision)...
1204
detections_count = 8813, unique_truth_count = 2997
class_id = 0, name = backpack, ap = 75.21% (TP = 765, FP = 173)
class_id = 1, name = handbag, ap = 44.04% (TP = 429, FP = 251)
class_id = 2, name = suitcase, ap = 73.82% (TP = 475, FP = 112)
for conf_thresh = 0.25, precision = 0.76, recall = 0.56, F1-score = 0.64
for conf_thresh = 0.25, TP = 1669, FP = 536, FN = 1328, average IoU = 60.23 %
IoU threshold = 50 %, used Area-Under-Curve for each unique Recall
mean average precision (mAP@0.50) = 0.643579, or 64.36 %
Total Detection Time: 15.000000 Seconds
Set -points flag:
`-points 101` for MS COCO
`-points 11` for PascalVOC 2007 (uncomment `difficult` in voc.data)
`-points 0` (AUC) for ImageNet, PascalVOC 2010-2012, your custom dataset
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$
6.3 network resolution = 480 * 480
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$ ./darknet detector map ./cfg/p*****_yolov3_bag.data ./cfg/p*****_yolov3_bag.cfg /media/famu/DISK_DATA/yongqiang/p*****_yolov3_bag_best.weights -i 3
net.optimized_memory = 0
batch = 1, time_steps = 1, train = 0
layer filters size/strd(dil) input output
0 conv 32 3 x 3/ 2 480 x 480 x 3 -> 240 x 240 x 32 0.100 BF
1 conv 16 1 x 1/ 1 240 x 240 x 32 -> 240 x 240 x 16 0.059 BF
2 conv 32 3 x 3/ 2 240 x 240 x 16 -> 120 x 120 x 32 0.133 BF
......
Total BFLOPS 9.395
Allocate additional workspace_size = 13.11 MB
Loading weights from /media/famu/DISK_DATA/yongqiang/p*****_yolov3_bag_best.weights...
seen 64
Done! Loaded 176 layers from weights-file
calculation mAP (mean average precision)...
1204
detections_count = 9370, unique_truth_count = 2997
class_id = 0, name = backpack, ap = 74.76% (TP = 746, FP = 162)
class_id = 1, name = handbag, ap = 41.37% (TP = 410, FP = 256)
class_id = 2, name = suitcase, ap = 73.20% (TP = 462, FP = 101)
for conf_thresh = 0.25, precision = 0.76, recall = 0.54, F1-score = 0.63
for conf_thresh = 0.25, TP = 1618, FP = 519, FN = 1379, average IoU = 60.07 %
IoU threshold = 50 %, used Area-Under-Curve for each unique Recall
mean average precision (mAP@0.50) = 0.631074, or 63.11 %
Total Detection Time: 14.000000 Seconds
Set -points flag:
`-points 101` for MS COCO
`-points 11` for PascalVOC 2007 (uncomment `difficult` in voc.data)
`-points 0` (AUC) for ImageNet, PascalVOC 2010-2012, your custom dataset
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$
6.4 network resolution = 416 * 416
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$ ./darknet detector map ./cfg/p*****_yolov3_bag.data ./cfg/p*****_yolov3_bag.cfg /media/famu/DISK_DATA/yongqiang/p*****_yolov3_bag_best.weights -i 3
net.optimized_memory = 0
batch = 1, time_steps = 1, train = 0
layer filters size/strd(dil) input output
0 conv 32 3 x 3/ 2 416 x 416 x 3 -> 208 x 208 x 32 0.075 BF
1 conv 16 1 x 1/ 1 208 x 208 x 32 -> 208 x 208 x 16 0.044 BF
2 conv 32 3 x 3/ 2 208 x 208 x 16 -> 104 x 104 x 32 0.100 BF
......
Total BFLOPS 7.057
Allocate additional workspace_size = 13.11 MB
Loading weights from /media/famu/DISK_DATA/yongqiang/p*****_yolov3_bag_best.weights...
seen 64
Done! Loaded 176 layers from weights-file
calculation mAP (mean average precision)...
1204
detections_count = 10236, unique_truth_count = 2997
class_id = 0, name = backpack, ap = 72.39% (TP = 726, FP = 180)
class_id = 1, name = handbag, ap = 37.99% (TP = 373, FP = 255)
class_id = 2, name = suitcase, ap = 70.34% (TP = 440, FP = 97)
for conf_thresh = 0.25, precision = 0.74, recall = 0.51, F1-score = 0.61
for conf_thresh = 0.25, TP = 1539, FP = 532, FN = 1458, average IoU = 58.44 %
IoU threshold = 50 %, used Area-Under-Curve for each unique Recall
mean average precision (mAP@0.50) = 0.602397, or 60.24 %
Total Detection Time: 13.000000 Seconds
Set -points flag:
`-points 101` for MS COCO
`-points 11` for PascalVOC 2007 (uncomment `difficult` in voc.data)
`-points 0` (AUC) for ImageNet, PascalVOC 2010-2012, your custom dataset
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$
6.5 network resolution = 352 * 352
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$ ./darknet detector map ./cfg/p*****_yolov3_bag.data ./cfg/p*****_yolov3_bag.cfg /media/famu/DISK_DATA/yongqiang/p*****_yolov3_bag_best.weights -i 3
net.optimized_memory = 0
batch = 1, time_steps = 1, train = 0
layer filters size/strd(dil) input output
0 conv 32 3 x 3/ 2 352 x 352 x 3 -> 176 x 176 x 32 0.054 BF
1 conv 16 1 x 1/ 1 176 x 176 x 32 -> 176 x 176 x 16 0.032 BF
2 conv 32 3 x 3/ 2 176 x 176 x 16 -> 88 x 88 x 32 0.071 BF
......
Total BFLOPS 5.052
Allocate additional workspace_size = 13.11 MB
Loading weights from /media/famu/DISK_DATA/yongqiang/p*****_yolov3_bag_best.weights...
seen 64
Done! Loaded 176 layers from weights-file
calculation mAP (mean average precision)...
1204
detections_count = 12055, unique_truth_count = 2997
class_id = 0, name = backpack, ap = 67.68% (TP = 671, FP = 200)
class_id = 1, name = handbag, ap = 32.12% (TP = 316, FP = 232)
class_id = 2, name = suitcase, ap = 63.89% (TP = 403, FP = 108)
for conf_thresh = 0.25, precision = 0.72, recall = 0.46, F1-score = 0.56
for conf_thresh = 0.25, TP = 1390, FP = 540, FN = 1607, average IoU = 56.13 %
IoU threshold = 50 %, used Area-Under-Curve for each unique Recall
mean average precision (mAP@0.50) = 0.545619, or 54.56 %
Total Detection Time: 12.000000 Seconds
Set -points flag:
`-points 101` for MS COCO
`-points 11` for PascalVOC 2007 (uncomment `difficult` in voc.data)
`-points 0` (AUC) for ImageNet, PascalVOC 2010-2012, your custom dataset
(base) yongqiang@famu-sys:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$
7. ./darknet detector test ./cfg/yolov3-tiny.data ./cfg/yolov3-tiny.cfg ./yolov3-tiny_best.weights -i 0 /home/strong/darknet_work/darknet_AlexeyAB_191221/darknet-master/data/person.jpg
strong@foreverstrong:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$ ./darknet detector test ./cfg/yolov3-tiny.data ./cfg/yolov3-tiny.cfg ./yolov3-tiny_best.weights -i 0 /home/strong/darknet_work/darknet_AlexeyAB_191221/darknet-master/data/person.jpg
strong@foreverstrong:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$ ./darknet detector test ./cfg/yolov3-tiny.data ./cfg/yolov3-tiny.cfg ./yolov3-tiny_best.weights -i 0 /home/strong/darknet_work/darknet_AlexeyAB_191221/darknet-master/data/person.jpg
net.optimized_memory = 0
batch = 1, time_steps = 1, train = 0
layer filters size/strd(dil) input output
0 conv 16 3 x 3/ 1 608 x 608 x 3 -> 608 x 608 x 16 0.319 BF
1 max 2x 2/ 2 608 x 608 x 16 -> 304 x 304 x 16 0.006 BF
2 conv 32 3 x 3/ 1 304 x 304 x 16 -> 304 x 304 x 32 0.852 BF
3 max 2x 2/ 2 304 x 304 x 32 -> 152 x 152 x 32 0.003 BF
4 conv 64 3 x 3/ 1 152 x 152 x 32 -> 152 x 152 x 64 0.852 BF
5 max 2x 2/ 2 152 x 152 x 64 -> 76 x 76 x 64 0.001 BF
6 conv 128 3 x 3/ 1 76 x 76 x 64 -> 76 x 76 x 128 0.852 BF
7 max 2x 2/ 2 76 x 76 x 128 -> 38 x 38 x 128 0.001 BF
8 conv 256 3 x 3/ 1 38 x 38 x 128 -> 38 x 38 x 256 0.852 BF
9 max 2x 2/ 2 38 x 38 x 256 -> 19 x 19 x 256 0.000 BF
10 conv 512 3 x 3/ 1 19 x 19 x 256 -> 19 x 19 x 512 0.852 BF
11 max 2x 2/ 1 19 x 19 x 512 -> 19 x 19 x 512 0.001 BF
12 conv 1024 3 x 3/ 1 19 x 19 x 512 -> 19 x 19 x1024 3.407 BF
13 conv 256 1 x 1/ 1 19 x 19 x1024 -> 19 x 19 x 256 0.189 BF
14 conv 512 3 x 3/ 1 19 x 19 x 256 -> 19 x 19 x 512 0.852 BF
15 conv 24 1 x 1/ 1 19 x 19 x 512 -> 19 x 19 x 24 0.009 BF
16 yolo
[yolo] params: iou loss: mse (2), iou_norm: 0.75, cls_norm: 1.00, scale_x_y: 1.00
17 route 13 -> 19 x 19 x 256
18 conv 128 1 x 1/ 1 19 x 19 x 256 -> 19 x 19 x 128 0.024 BF
19 upsample 2x 19 x 19 x 128 -> 38 x 38 x 128
20 route 19 8 -> 38 x 38 x 384
21 conv 256 3 x 3/ 1 38 x 38 x 384 -> 38 x 38 x 256 2.555 BF
22 conv 24 1 x 1/ 1 38 x 38 x 256 -> 38 x 38 x 24 0.018 BF
23 yolo
[yolo] params: iou loss: mse (2), iou_norm: 0.75, cls_norm: 1.00, scale_x_y: 1.00
Total BFLOPS 11.643
Allocate additional workspace_size = 52.43 MB
Loading weights from ./yolov3-tiny_best.weights...
seen 64
Done! Loaded 24 layers from weights-file
/home/strong/darknet_work/darknet_AlexeyAB_191221/darknet-master/data/person.jpg: Predicted in 5.491000 milli-seconds.
person: 88%
strong@foreverstrong:~/darknet_work/darknet_AlexeyAB_191221/darknet-master$
扫一扫
1827
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
