-
![image](https://user-images.githubusercontent.com/63961323/98797060-0def5a00-2447-11eb-9131-4b26b865a9bb.png)
-
我的数据只有一类行人,我用的是AB版Darknet,我只调了iou_normalizer参数,从0.07改成0.15还是一样,基本没啥变化。。我是从头开始train的。求助大神!以下是我的cfg文件中参数设置:
[net]
# Testing
#batch=1
#subdivisions=1
# Training
batch=64
subdivisions=16
width=416…
-
Hi!请问用rotated IOU loss是不是容易引起梯度爆炸?我在自定义的数据上表现出的梯度爆炸的情况,刚开始是没问题的,就是loss有点抖,但是后面就直接nan了;IOU loss应该不可导吧,请问这个Rotated IOU loss是怎么实现可微的,我看上交的严老师课题组基于Ratated IOU loss不可导的问题发展了两套近似的loss,比较好奇你们怎么做到可导的。期待您的答复~
-
May I ask what is going on? What causes it? This did not happen with YOLOv5
![image](https://user-images.githubusercontent.com/91411583/179669221-48f014f7-dcbc-4529-9d2a-e92a55342b89.png)
![image]…
-
下面是yolov4原论文中给出的方法,请问您的代码全部实现了这些功能吗?如果没有的话具体哪些还没有实现呢?谢谢!
YOLO v4 uses:
• Bag of Freebies (BoF) for backbone: CutMix and Mosaic data augmentation, DropBlock regularization,Class label smoothing
• …
TMRin updated
4 years ago
-
I would like to ask if Yolov4 still uses the previous detection approach stated in the original YOLO paper?
https://arxiv.org/pdf/1506.02640.pdf Specifically, on the Unified Detection.
With YOL…
-
非常感谢作者的贡献,还没有研究代码,先在看论文,遇到一些困惑,希望作者大佬能帮忙解答一下:
(1)objectness score是啥呢?文中有强调“Specifically, Dense Detector obtains objectness score by calculating the Complete Intersection over Union(CIoU) [38] between…
-
Thanks for your great work!
I train (Gaussian+GIOU+YOLOv3) in this repository.
Which command should I use to calculate mAP for (Gaussian+GIOU+YOLOv3) after finishing training in this repository,
I…
-
Hello,
I am writing a instance segmentation network and I adopt a SSD head as the detection head. If I want to adopt CIOU/DIOU, how should I modify my code to integrate them together? Can this repo…
-
### 🚀 The feature, motivation and pitch
I am trying to add a new loss function for using object detection time.
One Short Story: Suppose you create one project like face detection. And that time …