Enhancing Perceptual Loss with Adversarial Feature Matching for Super-Resolution
這篇文章提出的問題是感知損失中由於一開始vgg訓練的目的與超分目的不符，導致用感知損失約束的網絡會產生不符合原本分佈的HR僞影。 vgg一開始訓練的目的是爲了進行分類任務這導致在前面30層中利用conv進行特徵提取時對特徵很敏感，即輸入的特徵圖的某部分特徵會爲了去歸爲某一類而導致產生的特徵圖有不符合原分佈的特徵。 這篇文章提出的解決辦法爲利用對抗網絡中間層生成的特徵作損失約束網絡去除僞影，對抗網絡
相關文章
- 1. IMPROVING GENERATIVE ADVERSARIAL NETWORKS WITH DENOISING FEATURE MATCHING（Bingio-ICLR2017）
- 2. perceptual loss 和adversial loss
- 3. Adversarial learning with multi-scale loss for skin lesion segmentation
- 4. perceptual loss(感知loss)介紹
- 5. 2D Features framework7(Feature Matching with FLANN )
- 6. ExFuse- Enhancing Feature Fusion for Semantic Segmentation筆記
- 7. Enhancing Underwater Imagery using Generative Adversarial Networks
- 8. 【文獻閱讀】Perceptual Generative Adversarial Networks for Small Object Detection –CVPR-2017
- 9. Finding Tiny Faces in the Wild with Generative Adversarial Network
- 10. [論文筆記]FashionBERT: Text and Image Matching with Adaptive Loss for Cross-modal Retrieval
- 更多相關文章...
- • Swift for 循環 - Swift 教程
- • Scala for循環 - Scala教程
- • 爲了進字節跳動，我精選了29道Java經典算法題，帶詳細講解
- • 算法總結-股票買賣
相關標籤/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
歡迎關注本站公眾號,獲取更多信息
相關文章
- 1. IMPROVING GENERATIVE ADVERSARIAL NETWORKS WITH DENOISING FEATURE MATCHING（Bingio-ICLR2017）
- 2. perceptual loss 和adversial loss
- 3. Adversarial learning with multi-scale loss for skin lesion segmentation
- 4. perceptual loss(感知loss)介紹
- 5. 2D Features framework7(Feature Matching with FLANN )
- 6. ExFuse- Enhancing Feature Fusion for Semantic Segmentation筆記
- 7. Enhancing Underwater Imagery using Generative Adversarial Networks
- 8. 【文獻閱讀】Perceptual Generative Adversarial Networks for Small Object Detection –CVPR-2017
- 9. Finding Tiny Faces in the Wild with Generative Adversarial Network
- 10. [論文筆記]FashionBERT: Text and Image Matching with Adaptive Loss for Cross-modal Retrieval