Merge pull request #2 from sdcb/ir-queue-raw-ir

0.5.1
sdcb · Oct 31, 2023 · a9d344a · a9d344a
2 parents 15c180d + 9a21d70
commit a9d344a
Show file tree

Hide file tree

Showing 29 changed files with 639 additions and 182 deletions.
diff --git a/.editorconfig b/.editorconfig
@@ -0,0 +1,4 @@
+[*.cs]
+
+# xUnit1004: Test methods should not be skipped
+dotnet_diagnostic.xUnit1004.severity = silent
diff --git a/build/00-common.linq b/build/00-common.linq
@@ -21,9 +21,9 @@ static void DotNetRun(string args) => Run("dotnet", args.Dump(), Encoding.GetEnc
 static void Run(string exe, string args, Encoding encoding) => Util.Cmd(exe, args, encoding);
 static ProjectVersion[] Projects = new[]
 {
-	new ProjectVersion("Sdcb.OpenVINO", "0.4.5"),
-	new ProjectVersion("Sdcb.OpenVINO.Extensions.OpenCvSharp4", "0.1.0"),
-	new ProjectVersion("Sdcb.OpenVINO.PaddleOCR", "0.3.1"),
+	new ProjectVersion("Sdcb.OpenVINO", "0.5.1"),
+	new ProjectVersion("Sdcb.OpenVINO.Extensions.OpenCvSharp4", "0.2.0"),
+	new ProjectVersion("Sdcb.OpenVINO.PaddleOCR", "0.5.1"),
 	new ProjectVersion("Sdcb.OpenVINO.PaddleOCR.Models.Online", "0.2"),
 };
 

diff --git a/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR.Tests/CrashTest.cs b/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR.Tests/CrashTest.cs
@@ -112,10 +112,10 @@ await Task.WhenAll(Enumerable.Range(0, threadCount).Select(i => Task.Run(() =>
         })));
     }
 
-    [Fact(Skip = "Crash test only")]
+    [Fact(Skip = "Crash test only"), Obsolete("QueuedPaddleOcrAll")]
     public async Task QueuedTest()
     {
-        using Mat src = Cv2.ImRead("./samples/vsext.png"); 
+        using Mat src = Cv2.ImRead("./samples/vsext.png");
         //using Mat src = Cv2.ImDecode(await new HttpClient().GetByteArrayAsync("https://io.starworks.cc:88/paddlesharp/ocr/samples/xdr5450.webp"), ImreadModes.Color);
 
         FullOcrModel model = await OnlineFullModels.ChineseV4.DownloadAsync();
@@ -130,4 +130,18 @@ public async Task QueuedTest()
 
         await Task.WhenAll(Enumerable.Range(0, 100).Select(i => queued.Run(src)));
     }
+
+    [Fact]
+    public async Task OcrIsThreadSafe()
+    {
+        using Mat src = Cv2.ImRead("./samples/vsext.png"); 
+        using PaddleOcrAll ocr = new(await OnlineFullModels.ChineseV4.DownloadAsync());
+        Task[] tasks = Enumerable.Range(0, 2).Select(tidx => Task.Run(() => 
+        {
+            ocr.Run(src);
+            _console.WriteLine($"tid {tidx}: Good");
+        })).ToArray();
+
+        await Task.WhenAll(tasks);
+    }
 }
diff --git a/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR.Tests/OnlineClsTest.cs b/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR.Tests/OnlineClsTest.cs
@@ -37,7 +37,7 @@ public async Task ClsCPUBatch()
         ClsAsserts(cls, new[] { false, true }, new[] { src, src2 });
     }
 
-    [Fact]
+    [Fact(Skip = "GPU too slow")]
     public async Task ClsGpu()
     {
         using Mat src = Cv2.ImRead("./samples/5ghz.jpg");
@@ -48,7 +48,7 @@ public async Task ClsGpu()
         ClsAsserts(cls, true, src);
     }
 
-    [Fact]
+    [Fact(Skip = "GPU too slow")]
     public async Task ClsGpuBatch()
     {
         using Mat src = Cv2.ImRead("./samples/5ghz.jpg");
@@ -59,7 +59,7 @@ public async Task ClsGpuBatch()
         ClsAsserts(cls, new[] { false, true }, new[] { src, src2 });
     }
 
-    [Fact]
+    [Fact(Skip = "GPU too slow")]
     public async Task ClsGpuBatch10()
     {
         using Mat src = Cv2.ImRead("./samples/5ghz.jpg");

diff --git a/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR.Tests/OnlineModelsTest.cs b/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR.Tests/OnlineModelsTest.cs
@@ -39,12 +39,34 @@ public async Task FastCheckOCR()
                 _console.WriteLine("Detected all texts: \n" + result.Text);
                 foreach (PaddleOcrResultRegion region in result.Regions)
                 {
-                    _console.WriteLine($"Text: {region.Text}, Score: {region.Score}, RectCenter: {region.Rect.Center}, RectSize:    {region.Rect.Size}, Angle: {region.Rect.Angle}");
+                    _console.WriteLine($"Text: {region.Text}, Score: {region.Score}, RectCenter: {region.Rect.Center}, RectSize: {region.Rect.Size}, Angle: {region.Rect.Angle}");
                 }
             }
         }
     }
 
+    [Fact]
+    public async Task NoRotateTest()
+    {
+        FullOcrModel model = await OnlineFullModels.EnglishV3.DownloadAsync();
+
+        byte[] sampleImageData = File.ReadAllBytes(@"./samples/vsext.png");
+
+        using (PaddleOcrAll all = new(model)
+        {
+            AllowRotateDetection = false,
+            Enable180Classification = false,
+        })
+        {
+            using (Mat src = Cv2.ImDecode(sampleImageData, ImreadModes.Color))
+            {
+                PaddleOcrResult result = all.Run(src);
+                _console.WriteLine(result.Text);
+                Assert.Contains("IDE", result.Text);
+            }
+        }
+    }
+
     [Fact]
     public async Task FastCheckOCRWith180Cls()
     {
@@ -76,7 +98,7 @@ public async Task FastCheckOCRWith180Cls()
         }
     }
 
-    [Fact]
+    [Fact(Skip = "GPU too slow")]
     public async Task GPUFastCheckOCR()
     {
         FullOcrModel model = await OnlineFullModels.EnglishV3.DownloadAsync();

diff --git a/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR.Tests/OnlineRecTest.cs b/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR.Tests/OnlineRecTest.cs
@@ -60,7 +60,7 @@ public async Task RecCPUSmallerShape()
         RecAsserts(rec, src);
     }
 
-    [Fact]
+    [Fact(Skip = "GPU too slow")]
     public async Task RecGpu()
     {
         using Mat src = Cv2.ImRead("./samples/5ghz.jpg");
@@ -69,7 +69,7 @@ public async Task RecGpu()
         RecAsserts(rec, src);
     }
 
-    [Fact]
+    [Fact(Skip = "GPU too slow")]
     public async Task RecGpuLargerBatchShape()
     {
         using Mat src = Cv2.ImRead("./samples/5ghz.jpg");

diff --git a/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR/PaddleOcrClassifier.cs b/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR/PaddleOcrClassifier.cs
@@ -11,7 +11,7 @@ namespace Sdcb.OpenVINO.PaddleOCR;
 /// </summary>
 public class PaddleOcrClassifier : IDisposable
 {
-    private readonly InferRequest _p;
+    private readonly CompiledModel _compiledModel;
 
     /// <summary>
     /// Rotation threshold value used to determine if the image should be rotated.
@@ -36,10 +36,11 @@ public class PaddleOcrClassifier : IDisposable
     /// </summary>
     /// <param name="model">The <see cref="ClassificationModel"/> to use.</param>
     /// <param name="device">The device the inference request, pass null to using model's DefaultDevice.</param>
-    public PaddleOcrClassifier(ClassificationModel model, DeviceOptions? device = null)
+    public PaddleOcrClassifier(ClassificationModel model, 
+        DeviceOptions? device = null)
     {
         Shape = model.Shape;
-        _p = model.CreateInferRequest(device, prePostProcessing: (m, ppp) =>
+        _compiledModel = model.CreateCompiledModel(device, prePostProcessing: (m, ppp) =>
         {
             using PreProcessInputInfo ppii = ppp.Inputs.Primary;
             ppii.TensorInfo.Layout = Layout.NHWC;
@@ -50,7 +51,7 @@ public PaddleOcrClassifier(ClassificationModel model, DeviceOptions? device = nu
     /// <summary>
     /// Releases all resources used by the <see cref="PaddleOcrClassifier"/> object.
     /// </summary>
-    public void Dispose() => _p.Dispose();
+    public void Dispose() => _compiledModel.Dispose();
 
     /// <summary>
     /// Determines whether the image should be rotated by 180 degrees based on the threshold value.
@@ -103,23 +104,22 @@ private Ocr180DegreeClsResult[] BatchedShouldRotate180(Mat[] srcs)
     {
         using Mat final = PrepareAndStackImages(srcs);
 
+        using InferRequest ir = _compiledModel.CreateInferRequest();
         using (Tensor input = final.StackedAsTensor(srcs.Length))
         {
-            _p.Inputs.Primary = input;
-            _p.Run();
+            ir.Inputs.Primary = input;
+            ir.Run();
         }
 
-        using (Tensor output = _p.Outputs.Primary)
+        using Tensor output = ir.Outputs.Primary;
+        ReadOnlySpan<float> data = output.GetData<float>();
+        Ocr180DegreeClsResult[] results = new Ocr180DegreeClsResult[data.Length / 2];
+        for (int i = 0; i < results.Length; i++)
         {
-            ReadOnlySpan<float> data = output.GetData<float>();
-            Ocr180DegreeClsResult[] results = new Ocr180DegreeClsResult[data.Length / 2];
-            for (int i = 0; i < results.Length; i++)
-            {
-                results[i] = new Ocr180DegreeClsResult(data[(i * 2)..((i + 1) * 2)], RotateThreshold);
-            }
-
-            return results;
+            results[i] = new Ocr180DegreeClsResult(data[(i * 2)..((i + 1) * 2)], RotateThreshold);
         }
+
+        return results;
     }
 
     private Mat PrepareAndStackImages(Mat[] srcs)
@@ -135,7 +135,7 @@ private Mat PrepareAndStackImages(Mat[] srcs)
                     {
                         4 => src.CvtColor(ColorConversionCodes.RGBA2RGB),
                         1 => src.CvtColor(ColorConversionCodes.GRAY2RGB),
-                        3 => src.WeakRef(),
+                        3 => src.FastClone(),
                         var x => throw new Exception($"Unexpect src channel: {x}, allow: (1/3/4)")
                     };
                     return ResizePadding(channel3, Shape);
@@ -191,7 +191,7 @@ private static Mat ResizePadding(Mat src, NCHW shape)
         double whRatio = 1.0 * shape.Width / shape.Height;
         using Mat roi = 1.0 * srcSize.Width / srcSize.Height > whRatio ?
             src[0, srcSize.Height, 0, (int)Math.Floor(1.0 * srcSize.Height * whRatio)] :
-            src.WeakRef();
+            src.FastClone();
 
         double scaleRate = 1.0 * shape.Height / srcSize.Height;
         Mat resized = roi.Resize(new Size(Math.Floor(roi.Width * scaleRate), shape.Height));

diff --git a/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR/PaddleOcrDetector.cs b/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR/PaddleOcrDetector.cs
@@ -12,8 +12,7 @@ namespace Sdcb.OpenVINO.PaddleOCR;
 /// </summary>
 public class PaddleOcrDetector : IDisposable
 {
-    /// <summary>Holds an instance of the InferRequest class.</summary>
-    readonly InferRequest _p;
+    readonly CompiledModel _compiledModel;
 
     /// <summary>
     /// Gets or sets the maximum size for resizing the input image.
@@ -66,7 +65,9 @@ public class PaddleOcrDetector : IDisposable
     /// <para>If this property is null, network can work with image of any size and h/w ratio (dynamic).</para>
     /// <para>Otherwise, network works with fixed size image (static).</para>
     /// </param>
-    public PaddleOcrDetector(DetectionModel model, DeviceOptions? options = null, Size? staticShapeSize = null)
+    public PaddleOcrDetector(DetectionModel model, 
+        DeviceOptions? options = null, 
+        Size? staticShapeSize = null)
     {
         if (staticShapeSize != null)
         {
@@ -75,7 +76,7 @@ public PaddleOcrDetector(DetectionModel model, DeviceOptions? options = null, Si
                 32 * Math.Ceiling(1.0 * staticShapeSize.Value.Height / 32));
         }
 
-        _p = model.CreateInferRequest(options, afterReadModel: m =>
+        _compiledModel = model.CreateCompiledModel(options, afterReadModel: m =>
         {
             if (model.Version != ModelVersion.V4)
             {
@@ -104,7 +105,7 @@ public PaddleOcrDetector(DetectionModel model, DeviceOptions? options = null, Si
     /// </summary>
     public void Dispose()
     {
-        _p.Dispose();
+        _compiledModel.Dispose();
     }
 
     /// <summary>
@@ -221,14 +222,15 @@ public unsafe Mat RunRaw(Mat src, out Size resizedSize)
             normalized = Normalize(padded);
         }
 
+        using InferRequest ir = _compiledModel.CreateInferRequest();
         using (Mat _ = normalized)
         using (Tensor input = normalized.AsTensor())
         {
-            _p.Inputs.Primary = input;
-            _p.Run();
+            ir.Inputs.Primary = input;
+            ir.Run();
         }
 
-        using (Tensor output = _p.Outputs[0])
+        using (Tensor output = ir.Outputs[0])
         {
             Span<float> data = output.GetData<float>();
             NCHW shape = output.Shape.ToNCHW();
@@ -263,23 +265,23 @@ private static float GetScore(Point[] contour, Mat pred)
 
     private static Mat MatResize(Mat src, int? maxSize)
     {
-        if (maxSize == null) return src.WeakRef();
+        if (maxSize == null) return src.FastClone();
 
         Size size = src.Size();
         int longEdge = Math.Max(size.Width, size.Height);
         double scaleRate = 1.0 * maxSize.Value / longEdge;
 
         return scaleRate < 1.0 ?
             src.Resize(Size.Zero, scaleRate, scaleRate) :
-            src.WeakRef();
+            src.FastClone();
     }
 
     private static Mat MatResize(Mat src, Size maxSize)
     {
         Size srcSize = src.Size();
         if (srcSize == maxSize)
         {
-            return src.WeakRef();
+            return src.FastClone();
         }
 
         double scale = Math.Min(maxSize.Width / (double)srcSize.Width, maxSize.Height / (double)srcSize.Height);
@@ -289,7 +291,7 @@ private static Mat MatResize(Mat src, Size maxSize)
 
         if (scale == 1)
         {
-            return src.WeakRef();
+            return src.FastClone();
         }
         else
         {

diff --git a/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR/PaddleOcrRecognizer.cs b/projects/PaddleOCR/Sdcb.OpenVINO.PaddleOCR/PaddleOcrRecognizer.cs
@@ -1,10 +1,8 @@
 using OpenCvSharp;
-using Sdcb.OpenVINO.Natives;
 using Sdcb.OpenVINO.Extensions.OpenCvSharp4;
 using Sdcb.OpenVINO.PaddleOCR.Models;
 using System;
 using System.Linq;
-using System.Runtime.InteropServices;
 using System.Text;
 
 namespace Sdcb.OpenVINO.PaddleOCR;
@@ -14,7 +12,7 @@ namespace Sdcb.OpenVINO.PaddleOCR;
 /// </summary>
 public class PaddleOcrRecognizer : IDisposable
 {
-    private readonly InferRequest _p;
+    private readonly CompiledModel _compiledModel;
 
     /// <summary>
     /// Recognization model being used for OCR.
@@ -53,12 +51,14 @@ public class PaddleOcrRecognizer : IDisposable
     /// The value will be rounded to the nearest upper multiple of 32. This parameter is useful for models that require a fixed shape input.
     /// Pass `null` if the model supports dynamic input shape.
     /// </param>
-    public PaddleOcrRecognizer(RecognizationModel model, DeviceOptions? deviceOptions = null, int? staticShapeWidth = null)
+    public PaddleOcrRecognizer(RecognizationModel model, 
+        DeviceOptions? deviceOptions = null, 
+        int? staticShapeWidth = null)
     {
         Model = model;
         StaticShapeWidth = staticShapeWidth.HasValue ? (int)(32 * Math.Ceiling(1.0 * staticShapeWidth.Value / 32)) : null;
 
-        _p = model.CreateInferRequest(deviceOptions, prePostProcessing: (m, ppp) =>
+        _compiledModel = model.CreateCompiledModel(deviceOptions, prePostProcessing: (m, ppp) =>
         {
             using PreProcessInputInfo ppii = ppp.Inputs.Primary;
             ppii.TensorInfo.Layout = Layout.NHWC;
@@ -75,7 +75,7 @@ public PaddleOcrRecognizer(RecognizationModel model, DeviceOptions? deviceOption
     /// <summary>
     /// Releases all resources used by the current instance of the <see cref="PaddleOcrRecognizer"/> class.
     /// </summary>
-    public void Dispose() => _p.Dispose();
+    public void Dispose() => _compiledModel.Dispose();
 
     /// <summary>
     /// Run OCR recognition on multiple images in batches.
@@ -136,14 +136,14 @@ private unsafe PaddleOcrRecognizerResult[] BatchedRun(Mat[] srcs)
         }));
 
         using Mat final = PrepareAndStackImages(srcs, modelHeight, maxWidth);
-
+        using InferRequest ir = _compiledModel.CreateInferRequest();
         using (Tensor input = final.StackedAsTensor(srcs.Length))
         {
-            _p.Inputs.Primary = input;
-            _p.Run();
+            ir.Inputs.Primary = input;
+            ir.Run();
         }
-
-        using (Tensor output = _p.Outputs.Primary)
+            
+        using (Tensor output = ir.Outputs.Primary)
         {
             IntPtr dataPtr = output.DangerousGetDataPtr();
             int dataLength = (int)output.Size;
@@ -218,7 +218,7 @@ private static unsafe Mat PrepareAndStackImages(Mat[] srcs, int modelHeight, int
                     {
                         4 => src.CvtColor(ColorConversionCodes.RGBA2RGB),
                         1 => src.CvtColor(ColorConversionCodes.GRAY2RGB),
-                        3 => src.WeakRef(),
+                        3 => src.FastClone(),
                         var x => throw new Exception($"Unexpect src channel: {x}, allow: (1/3/4)")
                     };
                     return ResizePadding(channel3, modelHeight, maxWidth);