-
作者您好，先介绍一下我的实验情况：我用下载的权重文件和自己训练的权重分别进行推理，mAP分别是27.1和23.8，这导致最终oriented rcnn检测DOTA测试集结果差距较大。下载的权重文件最终得到的结果与论文数据33.31基本一致，而我自己训练的权重最终测试集mAP只有28.15。
所以想请教一下，DOTA数据集在裁剪清洗阶段您是怎样操作的？是否裁剪策略不同会影响训练效果？
-
### Description
Basically, my account has some kind of problem that causes the game to lag on any PC, not just mine. I asked a friend to test it on his computer, and he also experienced lag using m…
-
Hi Ankita, Thanks for you work!
Do we need to train the model ourselves to obtain the checkpoints for direct inference with the DOTA model, or are they yet to be released?
-
### Description
after the game dota lagged and didnt count me pts after clear win . Whats wrong
### Example Match ID (and possibly Timestamp)
_No response_
### Screenshots
_No response_
-
Hi,
Im using naive on windows with nekobox
I think something is wrong with naive and nekobox
whenever im using naive on TUN mode I cant use some programs like `discord` and `dota 2`
discord strug…
-
您好，我训练了STD with Oriented RCNN and HiViT-B，训练集使用ms_trainval（多尺度），验证集使用ss_trainval（单尺度），测试集使用ms_test，训练18epoch，samples_per_gpu=2，workers_per_gpu=4，以下为配置文件，问什么我最终的结果
COCO style result:
AP50: 0.81977185…
-
#### Your system information
* System information
https://gist.github.com/u83rmensch/1c46ebab44588ede8fcefa094d17d396
* Have you checked for system updates?:
Yes
* Are you using the lates…
-
#### Your system information
* Steam client version (build number or date): 1689034492
* Distribution (e.g. Ubuntu): probably doesn't matter, tested on Ubuntu 22.04
* Opted into Steam client beta…
-
### Description
today my dota crashed 10 times mid anything (in-game,on finding match, accepting match, during draft, mainmenu, custom game, lobby) why god??????????????
It has never happened to me …
-
你好作者！
请问RFD放在我的backbone中，跑dota数据集需要做预训练吗。
如果需要，大概要多大计算资源呢？