-
## 一言でいうと
Few-shotの画像認識で、クラス間の階層構造を利用した研究。クラス間に共通する上位クラスを認識できるようにすることで、分類クラス数が多くなってもメタ認識を活用できるようにする。上位クラスは、クラス名のWord2Vec上の類似度を使用する方法をとっている。
![image](https://user-images.githubusercontent.com/544…
-
Hi @laiyongkui1997 @AtmaHou,
Thank you for the great work, I have read all the three papers Fewshot slot tagging, FewJoint benchmark and Few_shot joint learning ID and SF. There are some questions th…
-
**Problem**
What do you want to do? What is blocking you?
I am reading an article and would like to discuss the possibility of implementing it in the prototypical network (https://arxiv.org/abs/18…
-
https://arxiv.org/abs/1706.05064
TMats updated
7 years ago
-
## Abstract
- Zero-shot anomaly detection(ZSAD)은 target dataset에 대한 training sample 없이 auxiliary data을 사용해서 학습된 모델을 필요로 한다.
- ZSAD는 학습 데이터가 충분치 못한 상황에서 중요하지만, 모델이 다양한 도메인에서의 anomalies에 generalize되는 …
sy00n updated
11 months ago
-
[paper](https://arxiv.org/pdf/2204.14198.pdf)
## TL;DR
- **I read this because.. :** #116 읽고 나서 읽고 싶어짐. 예전에 성현님이 소개해주셨는데 디테일 모름. GPT 열풍으로 요즘 다시 많이 언급됨.
- **task :** Vision Language Model in g…
-
Two variations:
- [ ] `magic-commit` looks at recent commits on repo for inspiration
- [ ] User supplies their own jinja templates
-
### Expected behavior
`dynamic_one_shot` works as a QNode transform.
### Actual behavior
```python
@partial(qml.dynamic_one_shot, postselect_mode="fill-shots")
@qml.qnode(qml.device("default.qubi…
-
Hi everyone! I am sorry that I just started this project and I am new to this topic. I am wondering where the code for supervised audio classification is. I just saw zero-shot learning. Thanks!
-
https://www.zjp97.top/tracking/38-siamtol/
A personal blog that shares computer vision papers