Closed che1007 closed 1 year ago
这是he成数据集,gt offsets可参考https://github.com/nie-lang/DeepImageStitching-1.0/blob/main/Dataset_gen/dataset.py
如果我没记错的话,在这篇论文中,我好像没有把分辨率固定到128128,不过最后跑出来的效果好像和128128的差不多
了解,謝謝回覆,另外想請問這個ground truth是怎麼獲得的
也可以通过这个代码获得,不过实际的label有周围的无效像素区域,论文中为了美观把这部分区域裁剪掉了
也可以通过这个代码获得: 是指執行這篇UnsupervisedDeepImageStitching reconstruction的interence.py所獲得的拼接圖像嗎?
因為我目前在看這條式子,不太清楚I_hat和I指的指哪邊
那個代碼指的是https://github.com/nie-lang/DeepImageStitching-1.0/blob/main/Dataset_gen/dataset.py ?
对
謝謝你用心地回覆
你好,我正在閱讀Learning edge-preserved image stitching from multi-scale deep homography,想請問ground truth offsets是怎麼來的?![Capture](https://user-images.githubusercontent.com/67545227/192971100-8a49fa0b-a586-437d-ad1e-254eb93bc4d1.PNG)