-
## Background
- swin base 모델로만 앙상블을 해본 결과, LB 점수가 0.005~0.01 정도밖에 오르질 않았음. (pspnet, upernet, PAN)
- 백본의 다양성을 높이면 앙상블 효과를 더 끌어올릴 수 있을 것 같음.
## Content
- hrnet + ocr (mmseg). epoch 50
- valid…
-
J'effectue une déclaration, à la page indicateur retour congé maternité lorsque je fais précédent pour retourner sur la page indicateur écart taux augmentations et à la page indicateur écart taux pro…
-
https://arxiv.org/abs/1804.07612
-
I have been trying to train mask rcnn on custom dataset but I am not getting good results on my val or test dataset.
I also tried applying augmentations but that didn't work either. (see [issue](http…
-
Thank you for your work. I tried to reproduce the results, but the results were not good. I only did simple data enhancement during training, and did not make the number of positive and negative sampl…
LauJN updated
3 years ago
-
* https://arxiv.org/abs/1811.09030v2
* https://github.com/jackryo/ricap
improve YoloV3 and prevent overfitting
-
Hello Sir,
Thanks for providing the code for packnet-sfm.
I am currently trying to train on custom Carla dataset. My yaml configuration file looks like:
```
checkpoint:
filepath: '/wor…
-
https://keras.io/guides/preprocessing_layers/
-
![image](https://user-images.githubusercontent.com/73097321/162659025-e99c15f1-b269-4b05-9835-af02516b7241.png)
I noticed that you used an expanded training dataset in the training anopic phase, and …
Llj-l updated
2 years ago
-
**Example User Feedback:**
Is 'Erase' basically random occlusion?
We need to make proper documentation on all the options we use and how they are basically applied to the data.