-
I'm training with the coco2017 dataset. When the training reaches the 26th epoch, the displayed LR = 0.0000
[2020-11-17 12:49:56,134]-[train.py line:223]: === Epoch:[ 26/50],step:[42640/44640],img…
-
When I train with different iou loss, diou and giou works fine, but ciou does not.
It blocks the training telling "Saving checkpoints for 0 into ..." and no more progess.
-
/darknet detector test cfg/coco.data cfg/yolov4.cfg
CUDA-version: 11000 (11000), cuDNN: 8.1.1, CUDNN_HALF=1, GPU count: 1
CUDNN_HALF=1
OpenCV version: 4.4.0
0 : compute_capability = 610, cu…
-
While testing on a local video file. I came across this error
"""
double free or corruption (!prev)
Aborted (core dumped)
"""
command used : python3 darknet_video.py --input videos/case1.mp4 --o…
-
## ❔Question
if n:
nt += n # cumulative targets
ps = pi[b, a, gj, gi] # prediction subset corresponding to targets
# Regression
t_box = ps[:,:4…
-
Great work by @argusswift
I forked this repo and did some modifications to suit my case better. (You can also check my repo :-) )But the training part remains mostly the same. And I came across prob…
-
请问在train.py下的Yolo_loss里面,在bounding box regression的时候,为什么没有用到CIoU呢,只是在筛选anchor的时候用到。此外,在tool/region_loss.py里面使用MSE进行bounding box regression。论文中不是提到用CIoU-loss吗?是因为作者公开的源码和论文内容不全都一致吗?(新手入门,如有不对,还请指教)
…
-
[2020-10-08 18:30:55,013]-[train.py line:103]:Train datasets number is : 6056
[2020-10-08 18:30:55,014]-[train.py line:106]: ======= start training ======
[2020-10-08 18:30:55,015]…
-
I replace the gious loss in FCOS by dious loss, found no improvement and also dious loss converge slower than gious loss.
-
[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
Total BFLOPS 127.232
avg_outputs = 10…