boki/poe2-bot
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

easyocr work

Browse source
This commit is contained in:
Boki 2026-02-12 09:29:10 -05:00
parent 9f208b0606
commit 735b6f7157
7 changed files with 121 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

26
tools/OcrDaemon/Daemon.cs
View file

@ -105,9 +105,10 @@ static class Daemon
private static object HandleDiffOcrPython(OcrHandler ocrHandler, PythonOcrBridge pythonBridge, Request request)
{
var sw = System.Diagnostics.Stopwatch.StartNew();
var p = request.Threshold > 0
? new DiffOcrParams { DiffThresh = request.Threshold }
: new DiffOcrParams();
// Use default params (same wide crop as Tesseract path).
// Background subtraction below eliminates stash items from the image.
var p = new DiffOcrParams();
if (request.Threshold > 0) p.DiffThresh = request.Threshold;
var cropResult = ocrHandler.DiffCrop(request, p);
if (cropResult == null)
@ -115,22 +116,29 @@ static class Daemon
var (cropped, refCropped, current, region) = cropResult.Value;
using var _current = current;
using var _refCropped = refCropped;
// Apply background subtraction to isolate tooltip text.
// This removes stash items and game world — only tooltip text remains.
// No upscale (upscale=1) to keep the image small for EasyOCR speed.
// Hard threshold (softThreshold=false) produces clean binary for OCR.
using var processed = ImagePreprocessor.PreprocessWithBackgroundSub(
cropped, refCropped, dimPercentile: 40, textThresh: 60, upscale: 1, softThreshold: false);
cropped.Dispose();
refCropped.Dispose();
var diffMs = sw.ElapsedMilliseconds;
// Save crop to requested path if provided
// Save processed crop if path provided
if (!string.IsNullOrEmpty(request.Path))
{
var dir = Path.GetDirectoryName(request.Path);
if (!string.IsNullOrEmpty(dir) && !Directory.Exists(dir))
Directory.CreateDirectory(dir);
cropped.Save(request.Path, ImageUtils.GetImageFormat(request.Path));
processed.Save(request.Path, ImageUtils.GetImageFormat(request.Path));
}
// Send crop to Python via base64 over pipe (no temp file I/O)
// Send processed image to Python OCR via base64
sw.Restart();
var ocrResult = pythonBridge.OcrFromBitmap(cropped, request.Engine!);
cropped.Dispose();
var ocrResult = pythonBridge.OcrFromBitmap(processed, request.Engine!);
var ocrMs = sw.ElapsedMilliseconds;
Console.Error.WriteLine($" diff-ocr-python: diff={diffMs}ms ocr={ocrMs}ms total={diffMs + ocrMs}ms crop={region.Width}x{region.Height}");