add ocr confidence logging and add filter threshold

CompatBot/EventHandlers/MediaScreenshotMonitor.cs

@@ -76,17 +76,18 @@ internal sealed class MediaScreenshotMonitor
 
             try
             {
-                if (await OcrProvider.GetTextAsync(item.imgUrl, Config.Cts.Token).ConfigureAwait(false) is {Length: >0} result
+                if (await OcrProvider.GetTextAsync(item.imgUrl, Config.Cts.Token).ConfigureAwait(false) is ({Length: >0} result, var confidence)
                     && !Config.Cts.Token.IsCancellationRequested)
                 {
                     var cnt = true;
                     var prefix = $"[{item.msg.Id % 100:00}]";
-                    var ocrTextBuf = new StringBuilder($"OCR result of message <{item.msg.JumpLink}>:").AppendLine();
-                    Config.Log.Debug($"{prefix} OCR result of message {item.msg.JumpLink}:");
+                    var ocrTextBuf = new StringBuilder($"OCR result of message <{item.msg.JumpLink}> ({confidence*100:0.00}%):").AppendLine();
+                    Config.Log.Debug($"{prefix} OCR result of message {item.msg.JumpLink} ({confidence*100:0.00}%):");
                     var duplicates = new HashSet<string>();
                     ocrTextBuf.AppendLine(result.Sanitize());
                     Config.Log.Debug($"{prefix} {result}");
                     if (cnt
+                        && confidence > 0.65
                         && await ContentFilter.FindTriggerAsync(FilterContext.Chat, result).ConfigureAwait(false) is Piracystring hit
                         && duplicates.Add(hit.String))
                     {

CompatBot/Ocr/Backend/AzureVision.cs

@@ -22,7 +22,7 @@ public class AzureVision: IOcrBackend
         return Task.FromResult(true);
     }
 
-    public async Task<string> GetTextAsync(string imgUrl,  CancellationToken cancellationToken)
+    public async Task<(string result, double confidence)> GetTextAsync(string imgUrl,  CancellationToken cancellationToken)
     {
         var headers = await cvClient.ReadAsync(imgUrl, cancellationToken: cancellationToken).ConfigureAwait(false);
         var operationId = new Guid(new Uri(headers.OperationLocation).Segments.Last());
@@ -43,10 +43,10 @@ public class AzureVision: IOcrBackend
                 foreach (var r in result.AnalyzeResult.ReadResults)
                 foreach (var l in r.Lines)
                     ocrTextBuf.AppendLine(l.Text);
-                return ocrTextBuf.ToString();
+                return (ocrTextBuf.ToString(), 1);
             }
         }
         Config.Log.Warn($"Failed to OCR image {imgUrl}: {result.Status}");
-        return "";
+        return ("", 0);
     }
 }

CompatBot/Ocr/Backend/BackendBase.cs

@@ -25,7 +25,7 @@ public abstract class BackendBase: IOcrBackend, IDisposable
         return Task.FromResult(true);
     }
 
-    public abstract Task<string> GetTextAsync(string imgUrl, CancellationToken cancellationToken);
+    public abstract Task<(string result, double confidence)> GetTextAsync(string imgUrl, CancellationToken cancellationToken);
 
     public virtual void Dispose() => HttpClient.Dispose();
 

CompatBot/Ocr/Backend/Florence2.cs

@@ -53,13 +53,13 @@ public class Florence2: BackendBase
         return true;
     }
 
-    public override async Task<string> GetTextAsync(string imgUrl, CancellationToken cancellationToken)
+    public override async Task<(string result, double confidence)> GetTextAsync(string imgUrl, CancellationToken cancellationToken)
     {
         await using var imgStream = await HttpClient.GetStreamAsync(imgUrl, cancellationToken).ConfigureAwait(false);
         var results = model.Run(TaskTypes.OCR_WITH_REGION, [imgStream], "", CancellationToken.None);
         var result = new StringBuilder();
         foreach (var box in results[0].OCRBBox)
             result.AppendLine(box.Text);
-        return result.ToString().TrimEnd();
+        return (result.ToString().TrimEnd(), 1);
     }
 }

CompatBot/Ocr/Backend/IOcrBackend.cs

@@ -4,5 +4,5 @@ public interface IOcrBackend
 {
     string Name { get; }
     Task<bool> InitializeAsync(CancellationToken cancellationToken);
-    Task<string> GetTextAsync(string imgUrl, CancellationToken cancellationToken);
+    Task<(string result, double confidence)> GetTextAsync(string imgUrl, CancellationToken cancellationToken);
 }

CompatBot/Ocr/Backend/Tesseract.cs

@@ -6,7 +6,7 @@ using TesseractCSharp.Interop;
 
 namespace CompatBot.Ocr.Backend;
 
-internal class Tesseract: BackendBase, IDisposable
+internal class Tesseract: BackendBase
 {
     private TesseractEngine engine;
 
@@ -65,12 +65,12 @@ internal class Tesseract: BackendBase, IDisposable
         return true;
     }
 
-    public override async Task<string> GetTextAsync(string imgUrl, CancellationToken cancellationToken)
+    public override async Task<(string result, double confidence)> GetTextAsync(string imgUrl, CancellationToken cancellationToken)
     {
         var imgData = await HttpClient.GetByteArrayAsync(imgUrl, cancellationToken).ConfigureAwait(false);
         using var img = Pix.LoadFromMemory(imgData);
         using var page = engine.Process(img);
-        return page.GetText() ?? "";
+        return (page.GetText() ?? "", page.GetMeanConfidence());
     }
 
     public override void Dispose()

CompatBot/Ocr/OcrProvider.cs

@@ -33,10 +33,11 @@ public static class OcrProvider
         }
     }
 
-    public static async Task<string> GetTextAsync(string imageUrl, CancellationToken cancellationToken)
+    public static async Task<(string result, double confidence)> GetTextAsync(string imageUrl, CancellationToken cancellationToken)
     {
         if (backend is null)
-            return "";
+            return ("", -1);
+
         try
         {
             return await backend.GetTextAsync(imageUrl, cancellationToken).ConfigureAwait(false);
@@ -44,7 +45,7 @@ public static class OcrProvider
         catch (Exception e)
         {
             Config.Log.Warn(e, $"Failed to OCR image {imageUrl}");
-            return "";
+            return ("", 0);
         }
     }