ソ瑠璃(soruly)

It seems that you have not created the .env file. Have you been following the guide in README?

I'll create a guide for linux when I've time. But windows 10 isn't something I'd like to support. I'm not interested in running docker with WSL2

> I'll create a guide for linux when I've time. But windows 10 isn't something I'd like to support. I'm not interested in running docker with WSL2 WSL2 is now...

The import function is useful but I want a more complete implementation of this feature. Like how it scan for hashes in file system? import from another server directly? how...

> Importing provided database dumps without media I would include a sql file in next database dump. So admin doesn't need to run another script to generate it. As you...

since revision https://github.com/soruly/trace.moe-api/commit/16b5847669a7d4a099b92618636e2d68634820cc , the worker process is now spawn directly in trace.moe-api. The rate limit would no longer affect hashing/loading, so bulk import mode is no longer needed. However,...

需要多少運算資源完全取決於你的資料量。 LIRE 很快的, 尤其是只選用比較簡單的 ColorLayout 時, 用 7950X3D 處理 17500 張圖片(320x180px)只需要 1-2 秒。 如果你只得一萬幾千張圖，隨便一台十年前的電腦應該也足夠應付。另外也盡可能確保有足夠的記憶體把整個資料庫放在 RAM 裡, 或者使用超快的 nvme SSD 也可以。LIRE 的 ColorLayout image hash 每張大約佔用 210 Bytes. 然而 ColorLayout 只計算...

頭像識別可以看看 https://github.com/nagadomi/animeface-2009 但我目前還沒有見過大規模使用的項目 色彩上的特徵明顯，是其中一個 ColorLayout 能找到匹配圖片的關鍵，若應用在比較灰暗的真實圖片，效果會比色彩更鮮明的動漫圖片差。ColorLayout 不會分析圖片上的邊緣，所以目前 trace.moe 為了提升 hash 速度和 deldup 效率，hash 的時候會把所有影格縮小至 320x180。這相比使用 720p 來說，搜尋準確度也沒有明顯分別。但若果使用邊緣相關的算法，例如 Edge Histogram 的話，解像度太低就會有很大影響。 關於 ACG 相關的項目可以參考一下 https://github.com/soruly/awesome-acg

trace.moe doesn't use python. I think it's not suitable to run on Colab.

trace.moe is not a simple program, the whole system consists of many parts. I think using docker-compose is already the easiest way to launch the whole thing. Some other developers...