CoSA: Correlation-Guided Change A ttention with Learnable Residual Gating for Remote Sensing Change Detection
Abdirashid Omar, Jonghyuk Park
実装難易度
Hard
推論・学習コスト
High
想定用途
検出
概要
Pixel-level annotation of fixed traffic-camera imagery is expensive, while crosswalk models trained from street-level imagery face a substantial viewpoint and appearance shift when applied to elevated CCTV. We investigate a data-efficient target-domain pipeline using 241 manually annotated CCTV images and 5,926 unlabeled CCTV frames. A source-domain experiment trains a 31.0M-parameter custom U-Net…
何が新しいか
Pixel-level annotation of fixed traffic-camera imagery is expensive, while crosswalk models trained from street-level imagery face a substantial viewpoint and appearance shift when applied to…
何に使えるか
検出
実装情報
- Paper URL
- あり
実装チェックリスト
実装または配布ページ
要確認Paper onlyの可能性があるため再実装前提で確認してください。
一次情報リンク
OKPaper
検証しやすさ
要確認公式実装が見つからないため、論文から再実装する前提です。
計算資源
要確認学習や高解像度推論ではGPUメモリと実行時間に注意が必要です。
ライセンス
未取得配布元のLICENSE、モデルカード、Paperの利用条件を確認してください。
商用利用
未取得研究利用限定、データセット由来制限、API規約の有無を確認してください。
自社データで試すなら
製造業・材料開発のExcel/CSVデータに落とし込むための最初の手順です。
- 1まず自社データを、入力条件、目的変数、評価したい指標に分けて整理します。
- 2正常データだけで動くベースラインを作り、異常スコアのしきい値を現場知見と合わせます。
- 3評価指標はR2/RMSE、AUC、異常検知の再現率、実験回数削減率など、現場の意思決定に近いものを選びます。
- 4SHAPや特徴量重要度で、効いている因子が物理・化学・工程知識と矛盾しないか確認します。
実装難易度
Hard - 公式実装が見つからないため、論文から再実装する前提です。
必要リソース
- GPU目安: High
- データセット: 論文・リポジトリ側の指定を確認してください。
- 学習要否: 再学習や評価環境の準備が必要になる可能性があります。
- 学習や高解像度推論ではGPUメモリと実行時間に注意が必要です。
実務で使う場合の注意点
- ライセンスと商用利用条件は、Paper / GitHub / Hugging Face の配布元で確認してください。
- 精度、再現性、計算コストはデータセットや評価条件に依存します。
- 個人情報や機密データを扱う場合は、入力データの保存先と外部API利用条件を確認してください。
関連記事
PICANet: Physics-Informed Cascaded Asymmetric Network for Infrared Small Target Detection
Infrared small target detection (ISTD) is an important research direction in image processing. However, existi
Weakly supervised neural network: segmentation of complex structures in X-ray microCT
Segmentation of complex structures in X-ray tomographic data is a fundamental task in biomedical research, but
WSPolypNet: Weakly Supervised Polyp Localization in Colonoscopy Videos
Because dense frame-level annotation of colonoscopy videos is costly, we propose WSPolypNet, a weakly supervis
Pre-Whitening and BCJR Posterior Distillation for Bi-LSTM Detection in Faster-than-Nyquist Signaling
Recurrent detectors such as bidirectional long short-term memory (Bi-LSTM) networks are low-complexity alterna