refactor(Avalonia): revamp OCR model download progress

It is now more transparent and more human.

Author: BrycensRanch

SnapX.Avalonia/Views/OCR.axaml.cs

@@ -12,6 +12,7 @@
 using FluentAvalonia.UI.Controls;
 using FluentAvalonia.UI.Windowing;
 using SixLabors.ImageSharp;
+using SixLabors.ImageSharp.Formats.Webp;
 using SixLabors.ImageSharp.PixelFormats;
 using SixLabors.ImageSharp.Processing;
 using SnapX.Avalonia.ViewModels;
@@ -134,7 +135,11 @@ private async Task RunOCRAsync(string languageCode, CancellationToken cts = defa
 
                 // Resize the image to match the Canvas exactly otherwise text boxes aren't where they should be.
                 response.AnnotatedImage.Mutate(x => x.Resize((int)TextOverlayCanvas.Width, (int)TextOverlayCanvas.Height, KnownResamplers.Bicubic));
-                await response.AnnotatedImage.SaveAsWebpAsync(ms, cts).ConfigureAwait(true);
+                await response.AnnotatedImage.SaveAsWebpAsync(ms, new WebpEncoder()
+                {
+                    Quality = 85
+                },
+                    cts).ConfigureAwait(true);
 
                 ms.Seek(0, SeekOrigin.Begin);
                 var bitmap = new Bitmap(ms);

SnapX.Core/Job/TaskHelpers.cs

@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 
+using System.Collections.Concurrent;
 using System.Diagnostics;
 using System.Diagnostics.CodeAnalysis;
 using System.Reflection;
@@ -992,126 +993,109 @@ private static OcrResponse MapToResponse(OcrResult result, Image visualImage)
 #endif
 
     public static async Task<OcrResponse> OCRImageDetailed(
-        Image? image = null,
-        string? filePath = null,
-        TaskSettings? taskSettings = null,
-        string? languageCode = null,
-        IProgress<OCRProgress>? progress = null,
-        CancellationToken cts = default
-    )
-
+    Image? image = null,
+    string? filePath = null,
+    TaskSettings? taskSettings = null,
+    string? languageCode = null,
+    IProgress<OCRProgress>? progress = null,
+    CancellationToken cts = default
+)
     {
 #if DISABLE_OCR
-DebugHelper.WriteException(
-
-new Exception("This build of SnapX was built with DISABLE_OCR build time constant.")
-
-);
-
-return new OcrResponse { FullText = "OCR Disabled in this build." };
-
+    DebugHelper.WriteException(new Exception("This build of SnapX was built with DISABLE_OCR build time constant."));
+    return new OcrResponse { FullText = "OCR Disabled in this build." };
 #else
-
-        progress?.Report(
-            new OCRProgress(
-                10,
-                $"Loading model for {languageCode}... (This may take awhile, I MAY download model files from internet)"
-            )
-        );
-
-        await Task.Yield();
-
-        progress?.Report(new OCRProgress(40, "Running OCR inference..."));
-
-        await Task.Yield();
-
+        progress?.Report(new(5, "Initializing OCR engine..."));
 
         var imageConfig = Configuration.Default;
-
         imageConfig.ImageFormatsManager.SetEncoder(AVIFFormat.Instance, AVIFEncoder.Instance);
-
         imageConfig.ImageFormatsManager.SetDecoder(AVIFFormat.Instance, AVIFDecoder.Instance);
-
-        imageConfig.ImageFormatsManager.AddImageFormatDetector(
-            new PatchedAVIFImageFormatDetector()
-        );
-
+        imageConfig.ImageFormatsManager.AddImageFormatDetector(new PatchedAVIFImageFormatDetector());
 
         try
-
         {
             if (filePath is not null && image is null)
-
             {
-                progress?.Report(new OCRProgress(45, "Reading image from disk..."));
-
-                await Task.Yield();
-
+                progress?.Report(new(10, "Reading image from disk..."));
                 image = await Image.LoadAsync(filePath, cts);
             }
         }
-
         catch (Exception ex)
-
         {
             var issue = "Failed to load image for OCR";
-
             DebugHelper.Logger?.Warning(issue);
-
             DebugHelper.WriteException(ex);
-
-            return new OcrResponse
-            {
-                FullText = issue + Environment.NewLine + ex.Message
-            };
+            return new OcrResponse { FullText = $"{issue}{Environment.NewLine}{ex.Message}" };
         }
 
-        if (image is null)
-
-            return new OcrResponse
-            {
-                FullText = "SNAPX ERROR: PASSED NULL IMAGE AND NULL FILEPATH."
-            };
+        if (image is null) return new OcrResponse { FullText = "SNAPX ERROR: PASSED NULL IMAGE AND NULL FILEPATH." };
 
-        progress?.Report(new OCRProgress(55, "Preprocessing image for Paddle..."));
+        var modelDir = Path.Combine(BaseDirectory.CacheHome, SnapX.AppName, "PaddleOCRModels");
+        var model = await GetModelForLanguage(languageCode ?? "eng");
+        var ocrEngine = new RapidOcr();
 
-        await Task.Yield();
+        var progressValue = 15;
+        var fileUrls = new ConcurrentDictionary<string, int>();
+        var fileProgressValues = new ConcurrentDictionary<string, int>();
 
-        DebugHelper.WriteLine(filePath);
+        EventHandler<HttpProgressEventArgs> progressHandler = async (sender, args) =>
+        {
+            var url = args.UserState as string;
+            if (string.IsNullOrEmpty(url)) return;
+            if (!url.Contains("PP-OCR", StringComparison.OrdinalIgnoreCase) &&
+                !url.Contains("ppocr", StringComparison.OrdinalIgnoreCase)) return;
 
+            fileUrls.TryAdd(url, fileUrls.Count);
 
-        progress?.Report(new OCRProgress(65, "Initializing Paddle Inference device..."));
+            var fileIndex = fileUrls[url];
+            var fileCount = Math.Max(1, fileUrls.Count);
+            var perFileRange = 70 / fileCount;
 
-        await Task.Yield();
+            var fileStartPct = 15 + (fileIndex * perFileRange);
+            var fileEndPct = fileStartPct + perFileRange;
 
-        DebugHelper.WriteLine($"OCR image bytes: ??");
+            var currentFileProgress = fileStartPct + (int)((args.ProgressPercentage / 100.0) * (fileEndPct - fileStartPct));
+            var lastFileProgress = fileProgressValues.GetOrAdd(url, 0);
 
-        progress?.Report(new OCRProgress(75, "Performing Text Detection & Recognition..."));
+            var uri = new Uri(url);
+            var query = System.Web.HttpUtility.ParseQueryString(uri.Query);
+            var filename = query["filename"] ?? query["name"] ?? query["response-content-disposition"];
 
-        await Task.Yield();
+            if (string.IsNullOrEmpty(filename)) filename = Path.GetFileName(uri.LocalPath);
+            if (!string.IsNullOrEmpty(filename) && filename.Contains("filename="))
+                filename = filename.Split("filename=").Last().Trim('"');
 
-        var ocrEngine = new RapidOcr();
-        var modelDir = Path.Combine(
+            filename ??= "model.onnx";
 
-            BaseDirectory.CacheHome,
+            var displayProgress = Math.Max(progressValue, currentFileProgress);
+            progress?.Report(new(displayProgress, $"Downloading {filename} ({args.ProgressPercentage}%)..."));
 
-            SnapX.AppName,
+            if (currentFileProgress > lastFileProgress)
+            {
+                fileProgressValues[url] = currentFileProgress;
+                if (currentFileProgress > progressValue) progressValue = currentFileProgress;
+            }
+            await Task.Yield();
+        };
 
-            "PaddleOCRModels"
+        if (HttpClientFactory._ph != null) HttpClientFactory._ph.HttpReceiveProgress += progressHandler;
 
-            );
-        var model = await GetModelForLanguage(languageCode ?? "eng");
+        try
+        {
+            progress?.Report(new(15, "Checking and loading OCR models..."));
+            await ocrEngine.LoadModelAsync(model, modelDir, HttpClientFactory.Get(), ct: cts);
+        }
+        finally
+        {
+            if (HttpClientFactory._ph != null) HttpClientFactory._ph.HttpReceiveProgress -= progressHandler;
+        }
 
-        await ocrEngine.LoadModelAsync(model, modelDir, HttpClientFactory.Get(), ct: cts);
+        progress?.Report(new(85, "Preprocessing image for Paddle..."));
         var originalLongSide = Math.Max(image.Width, image.Height);
         var targetLimit = 1504;
+        var finalResize = originalLongSide < targetLimit ? (int)(Math.Ceiling(originalLongSide / 32.0) * 32) : targetLimit;
 
-        // This logic ensures we don't upscale tiny images,
-        // but we still snap the result to a multiple of 32.
-        var finalResize = originalLongSide < targetLimit
-            ? (int)(Math.Ceiling(originalLongSide / 32.0) * 32)
-            : targetLimit;
-
+        progress?.Report(new(90, "Performing Text Detection & Recognition..."));
         var ocrResult = ocrEngine.Detect(image, new RapidOcrOptions
         {
             BoxScoreThresh = 0.5f,
@@ -1120,31 +1104,17 @@ public static async Task<OcrResponse> OCRImageDetailed(
             MostAngle = true,
             UnClipRatio = 1.6f,
             ImgResize = finalResize,
-            Padding = 10, // Reduced to keep text from getting lost in margins
+            Padding = 10,
         });
-        var visualDebugImage = image.Clone(_ => { });
-
 
+        var visualDebugImage = image.Clone(_ => { });
         foreach (var block in ocrResult.TextBlocks)
-
         {
-            DebugHelper.WriteLine(block.ToString());
-
-            var points = block.BoxPoints;
-
-
-            visualDebugImage.Mutate(ctx => { ctx.DrawPolygon(Color.Red, 4f, points); });
+            visualDebugImage.Mutate(ctx => { ctx.DrawPolygon(Color.Red, 4f, block.BoxPoints); });
         }
 
-
-        progress?.Report(new OCRProgress(95, "Finalizing results..."));
-
-        await Task.Yield();
-
-        DebugHelper.WriteAlways("Detected all texts: \n" + ocrResult.StrRes);
-
+        progress?.Report(new(98, "Finalizing results..."));
         return MapToResponse(ocrResult, visualDebugImage);
-
 #endif
     }
 

SnapX.Core/SnapX.cs

@@ -486,18 +486,7 @@ public static void InitSentry(Telemetry telemetry)
                 options.Environment = "production";
                 options.CreateHttpMessageHandler = () => HttpClientFactory.Handler;
 #endif
-                  options.ConfigureClient = client =>
-                  {
-                      var snapXHttpClient = HttpClientFactory.Get();
-
-                      foreach (var header in snapXHttpClient.DefaultRequestHeaders)
-                      {
-                          client.DefaultRequestHeaders.TryAddWithoutValidation(header.Key, header.Value);
-                      }
-
-                      client.DefaultRequestVersion = snapXHttpClient.DefaultRequestVersion;
-                      client.DefaultVersionPolicy = snapXHttpClient.DefaultVersionPolicy;
-                  };
+                  options.ConfigureClient = HttpClientFactory.ConfigureClient;
                   // VLCException includes multiple paths with username
                   // For full transparency, I discovered this issue on my computer.
                   // No other users are effected to my knowledge.