thtrieu/darkflow

Can we Do a transfert learning with a new label ?

Open

#117 建立於 2017年3月30日

在 GitHub 查看
 (8 留言) (5 反應) (0 負責人)Python (6,120 star) (2,102 fork)batch import
enhancementhelp wanted

描述

Hello

I had trained the model with 2 classes, and I want to use the generated .ckpt, to train the model woth news classes, How can I do that? , because when I test it with a new configuration, the program didn't work

Thanks

貢獻者指南

技術棧
pythontensorflow
領域
machine learning
議題類型
research
難度面向新貢獻者的預計實作難度，1 表示很小改動，5 表示專家級工作。
3
預計時間有經驗貢獻者完成調查、實作、測試並準備 pull request 的粗略時間範圍。
half day
活動狀態議題目前的可參與程度：新鮮、活躍、陳舊、阻塞或等待維護者輸入。
stale
清晰度議題是否清楚說明預期改動、驗收標準和下一步。
needs investigation
前置要求
Python basicsTensorFlow basicsdarkflow usage
新手友善度1-100 的估計分數，表示該議題對首次貢獻者的友善程度。
30
研究方向
Examine the darkflow source code and documentation to understand how to load a pre trained checkpoint and retrain with new classes. Look at the training script (e.g., train.py) and config files for options like 'load' and 'train'. Search for similar issues or discussions in the repository to find if this feature is supported or requires modification.
Can we Do a transfert learning with a new label ? · thtrieu/darkflow#117 | Good First Issue