-
-
https://github.com/bepoetree/MTTOD/blob/7256673b083448c8ac2fdd8bb95cdb9baf52b2b1/reader.py#L121
I found that in the original code, one test example(num_turn=18) will be discarded. Simply change the a…
-
### 本周工作
1.上周，老师给我提出了要把血管分割网络应用于其他未标注数据集的问题。本周通过阅读了一些论文还有和周雅婷同学的探讨，把这个问题归结为域适应问题(domain adaption),属于迁移学习领域。域适应问题认为，两个不同的数据集具有不同的域(domain)(我对域的理解为图像数据的分布)。现假定A数据集有标注，而B数据集没有。在A数据集上训练出的网络在B数据集中进行测试得到的结…
-
Мини блог о курсовой работе.
-
I have pre-trained the BERT MLM on my domain-specific corpus. I learned that naive averaging the embeddings in a sentence is not good for downstream tasks, so I wonder if I may use the sentence-BERT s…
-
Hi ,
I have doubt regarding number of classes for mIou , as per paper there are 5 classes Imp surface , building , Low veg , car , Tree and mIou is considered for these 5 classes ,
But in Code ,…
-
Hello, I am running the original baseline from KITTI to cityscape. When I use ANCHOR_SCALES=[8,16,32], I get similar results in your paper, but when I adjust ANCHOR_SCALES to [4, 8, 16,32], I got a ve…
-
Hello sir ,I read your paper it was so intresting , so may I ask you to share its code
or if you can't may I ask you to help me with one part (Good Anti-Noise)
Thank you
-
I want to use other object detection data sets for this experiment, can I do this, such as Pascal VOC,thanks!
-
Hi author of MOTSYNTH
Can we use real data (e.ge. MOT17) with un-labeled and use unsupervised learning?