-
## タイトル: プロンプト駆動型時間領域適応による夜間UAV追跡
## リンク: https://arxiv.org/abs/2409.18533
## 概要:
低照度環境下での夜間UAV追跡は、ドメイン適応(DA)によって大きな進歩を遂げました。しかし、従来のDAトレーニングベースの手法は、UAVトラッカーの時間的コンテキストの差異を狭めるには不十分でした。この問題に対処するため、本研…
-
Every Breath You Don't Take: Deepfake Speech Detection Using Breath
https://arxiv.org/abs/2404.15143
-
This should be generalization of #23.
GitHub link is just a single type of link.
Bot should support multiple links.
There should be a whitelist of supported domains to prevent users from addition o…
-
## タイトル: CAR: コントラスト不変潜在正則化を用いたコントラスト非依存変形医用画像レジストレーション
## リンク: https://arxiv.org/abs/2408.05341
## 概要:
マルチコントラスト画像レジストレーションは、異なる画像コントラスト間における複雑な強度関係のため、困難なタスクです。従来の画像レジストレーション手法は、一般に入力画像ペアごとに反復的な…
-
See this article:
Coykendall, Jim, and Muhammad Zafrullah. "AP-domains and unique factorization." Journal of Pure and Applied Algebra 189.1-3 (2004): 27-35.
https://www.sciencedirect.com/science/a…
-
## タイトル: CROCODILE:対照的脱結合学習による因果関係に基づいたロバスト性の向上
## リンク: https://arxiv.org/abs/2408.04949
## 概要:
ドメインシフトのため、深層学習を用いた画像分類器は、学習時とは異なるドメインに適用されると、その性能が大幅に低下します。例えば、ある病院の胸部X線画像で学習した分類器は、スキャナーの設定や患者の特性の…
-
-
## タイトル: 視覚言語モデル適応における誤整合を因果関係の観点から再考する
## リンク: https://arxiv.org/abs/2410.12816
## 概要:
CLIPのような基盤となる視覚言語モデルは、下流タスクにおいて素晴らしい汎化能力を示しています。しかし、CLIPは特定のタスクに適応する際に、タスクのずれとデータのずれという2つのレベルの不整合問題を抱えています。ソ…
-
I am curious that i can attach different types of llama adapter on freezed model.
For example, Once I train 2 different Llama-adapters. (one llama adapter for legal words the other for one medical wo…
-
## タイトル: アノテーションのないプライベートな3D医用画像のセグメンテーションのためのドメイン間分布調整
## リンク: https://arxiv.org/abs/2410.09210
## 概要:
3D医用画像のセグメンテーション作業における手動アノテーションは、退屈で時間のかかる作業です。さらに、データプライバシーの問題は、医療分野におけるデータアノテーションの実施に対するクラ…