-
Notifications
You must be signed in to change notification settings - Fork 267
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add TFLite + Transformer example #179
Merged
+680
−5
Merged
Changes from all commits
Commits
File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
|
@@ -51,3 +51,6 @@ docs-gen | |
site | ||
*.bak | ||
.idea/appInsightsSettings.xml | ||
|
||
#TFLite | ||
*.tflite |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,28 @@ | ||
/* | ||
* Copyright 2024 The Android Open Source Project | ||
* | ||
* Licensed under the Apache License, Version 2.0 (the "License"); | ||
* you may not use this file except in compliance with the License. | ||
* You may obtain a copy of the License at | ||
* | ||
* https://www.apache.org/licenses/LICENSE-2.0 | ||
* | ||
* Unless required by applicable law or agreed to in writing, software | ||
* distributed under the License is distributed on an "AS IS" BASIS, | ||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
* See the License for the specific language governing permissions and | ||
* limitations under the License. | ||
*/ | ||
tasks.register('downloadModelFile1', Download) { | ||
src 'https://storage.googleapis.com/download.tensorflow.org/models/tflite/task_library/style_transfer/android/magenta_arbitrary-image-stylization-v1-256_fp16_prediction_1.tflite' | ||
dest project.ext.ASSET_DIR + '/predict_float16.tflite' | ||
overwrite false | ||
} | ||
|
||
tasks.register('downloadModelFile2', Download) { | ||
src 'https://storage.googleapis.com/download.tensorflow.org/models/tflite/task_library/style_transfer/android/magenta_arbitrary-image-stylization-v1-256_fp16_transfer_1.tflite' | ||
dest project.ext.ASSET_DIR + '/transfer_float16.tflite' | ||
overwrite false | ||
} | ||
|
||
preBuild.dependsOn downloadModelFile1, downloadModelFile2 |
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
230 changes: 230 additions & 0 deletions
230
...edia/video/src/main/java/com/example/platform/media/video/StyleTransferShaderProgram.java
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,230 @@ | ||
/* | ||
* Copyright 2024 The Android Open Source Project | ||
* | ||
* Licensed under the Apache License, Version 2.0 (the "License"); | ||
* you may not use this file except in compliance with the License. | ||
* You may obtain a copy of the License at | ||
* | ||
* https://www.apache.org/licenses/LICENSE-2.0 | ||
* | ||
* Unless required by applicable law or agreed to in writing, software | ||
* distributed under the License is distributed on an "AS IS" BASIS, | ||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
* See the License for the specific language governing permissions and | ||
* limitations under the License. | ||
*/ | ||
|
||
package com.example.platform.media.video; | ||
|
||
import android.content.Context; | ||
import android.graphics.Bitmap; | ||
import android.graphics.BitmapFactory; | ||
import android.opengl.GLES20; | ||
import android.opengl.GLUtils; | ||
import android.util.Log; | ||
|
||
import androidx.media3.common.VideoFrameProcessingException; | ||
import androidx.media3.common.util.GlProgram; | ||
import androidx.media3.common.util.GlUtil; | ||
import androidx.media3.common.util.Size; | ||
import androidx.media3.common.util.UnstableApi; | ||
import androidx.media3.effect.BaseGlShaderProgram; | ||
|
||
import com.google.common.collect.ImmutableMap; | ||
|
||
import org.tensorflow.lite.DataType; | ||
import org.tensorflow.lite.Interpreter; | ||
import org.tensorflow.lite.InterpreterApi; | ||
import org.tensorflow.lite.gpu.CompatibilityList; | ||
import org.tensorflow.lite.gpu.GpuDelegate; | ||
import org.tensorflow.lite.support.common.FileUtil; | ||
import org.tensorflow.lite.support.common.ops.DequantizeOp; | ||
import org.tensorflow.lite.support.common.ops.NormalizeOp; | ||
import org.tensorflow.lite.support.image.ImageProcessor; | ||
import org.tensorflow.lite.support.image.TensorImage; | ||
import org.tensorflow.lite.support.image.ops.ResizeOp; | ||
import org.tensorflow.lite.support.image.ops.ResizeWithCropOrPadOp; | ||
import org.tensorflow.lite.support.tensorbuffer.TensorBuffer; | ||
|
||
import java.io.IOException; | ||
import java.io.InputStream; | ||
import java.nio.ByteBuffer; | ||
|
||
import javax.microedition.khronos.opengles.GL10; | ||
|
||
// TODO: Migrate this class to Kotlin | ||
@UnstableApi | ||
final class StyleTransferShaderProgram extends BaseGlShaderProgram { | ||
|
||
private static final String TAG = "StyleTransferSP"; | ||
private static final String VERTEX_SHADER_PATH = "shaders/vertex_shader_transformation_es2.glsl"; | ||
private static final String FRAGMENT_SHADER_PATH = "shaders/fragment_shader_copy_es2.glsl"; | ||
|
||
private final GlProgram glProgram; | ||
private final InterpreterApi transformInterpreter; | ||
private final int inputTransformTargetHeight; | ||
private final int inputTransformTargetWidth; | ||
private final int[] outputTransformShape; | ||
|
||
private final TensorBuffer predictOutput; | ||
|
||
private int width; | ||
private int height; | ||
|
||
public StyleTransferShaderProgram(Context context, String styleAssetFileName) | ||
throws VideoFrameProcessingException { | ||
super(/* useHighPrecisionColorComponents= */ false, /* texturePoolCapacity= */ 1); | ||
|
||
try { | ||
glProgram = new GlProgram(context, VERTEX_SHADER_PATH, FRAGMENT_SHADER_PATH); | ||
|
||
Interpreter.Options options = new Interpreter.Options(); | ||
|
||
CompatibilityList compatibilityList = new CompatibilityList(); | ||
if (compatibilityList.isDelegateSupportedOnThisDevice()) { | ||
GpuDelegate.Options gpuDelegateOptions = compatibilityList.getBestOptionsForThisDevice(); | ||
GpuDelegate gpuDelegate = new GpuDelegate(gpuDelegateOptions); | ||
options.addDelegate(gpuDelegate); | ||
} else { | ||
options.setNumThreads(6); | ||
} | ||
String predictModel = "predict_float16.tflite"; | ||
String transferModel = "transfer_float16.tflite"; | ||
Interpreter predictInterpeter = | ||
new Interpreter(FileUtil.loadMappedFile(context, predictModel), options); | ||
transformInterpreter = | ||
InterpreterApi.create(FileUtil.loadMappedFile(context, transferModel), options); | ||
int inputPredictTargetHeight = predictInterpeter.getInputTensor(0).shape()[1]; | ||
int inputPredictTargetWidth = predictInterpeter.getInputTensor(0).shape()[2]; | ||
int[] outputPredictShape = predictInterpeter.getOutputTensor(0).shape(); | ||
|
||
inputTransformTargetHeight = transformInterpreter.getInputTensor(0).shape()[1]; | ||
inputTransformTargetWidth = transformInterpreter.getInputTensor(0).shape()[2]; | ||
outputTransformShape = transformInterpreter.getOutputTensor(0).shape(); | ||
|
||
InputStream inputStream = context.getAssets().open(styleAssetFileName); | ||
Bitmap styleImage = BitmapFactory.decodeStream(inputStream); | ||
inputStream.close(); | ||
TensorImage styleTensorImage = | ||
getScaledTensorImage(styleImage, inputPredictTargetWidth, inputPredictTargetHeight); | ||
predictOutput = TensorBuffer.createFixedSize(outputPredictShape, DataType.FLOAT32); | ||
predictInterpeter.run(styleTensorImage.getBuffer(), predictOutput.getBuffer()); | ||
} catch (IOException | GlUtil.GlException e) { | ||
Log.w(TAG, "Error setting up TfShaderProgram", e); | ||
throw new VideoFrameProcessingException(e); | ||
} | ||
} | ||
|
||
@Override | ||
public Size configure(int inputWidth, int inputHeight) { | ||
width = inputWidth; | ||
height = inputHeight; | ||
return new Size(inputWidth, inputHeight); | ||
} | ||
|
||
@Override | ||
public void drawFrame(int inputTexId, long presentationTimeUs) | ||
throws VideoFrameProcessingException { | ||
ByteBuffer pixelBuffer = ByteBuffer.allocateDirect(width * height * 4); | ||
|
||
Bitmap bitmap; | ||
int texId; | ||
try { | ||
int[] boundFramebuffer = new int[1]; | ||
GLES20.glGetIntegerv(GLES20.GL_FRAMEBUFFER_BINDING, boundFramebuffer, /* offset= */ 0); | ||
|
||
int fboId = GlUtil.createFboForTexture(inputTexId); | ||
GlUtil.focusFramebufferUsingCurrentContext(fboId, width, height); | ||
GLES20.glReadPixels( | ||
/* x= */ 0, | ||
/* y= */ 0, | ||
width, | ||
height, | ||
GLES20.GL_RGBA, | ||
GLES20.GL_UNSIGNED_BYTE, | ||
pixelBuffer); | ||
GlUtil.checkGlError(); | ||
bitmap = Bitmap.createBitmap(width, height, Bitmap.Config.ARGB_8888); | ||
bitmap.copyPixelsFromBuffer(pixelBuffer); | ||
|
||
Log.w(TAG, "Process frame at " + (presentationTimeUs / 1000) + " ms"); | ||
long before = System.currentTimeMillis(); | ||
TensorImage tensorImage = | ||
getScaledTensorImage(bitmap, inputTransformTargetWidth, inputTransformTargetHeight); | ||
Log.w(TAG, "- Scale " + (System.currentTimeMillis() - before) + " ms"); | ||
TensorBuffer outputImage = | ||
TensorBuffer.createFixedSize(outputTransformShape, DataType.FLOAT32); | ||
|
||
before = System.currentTimeMillis(); | ||
transformInterpreter.runForMultipleInputsOutputs( | ||
new Object[] {tensorImage.getBuffer(), predictOutput.getBuffer()}, | ||
ImmutableMap.<Integer, Object>builder().put(0, outputImage.getBuffer()).build()); | ||
|
||
Log.w(TAG, "- Run " + (System.currentTimeMillis() - before) + " ms"); | ||
|
||
before = System.currentTimeMillis(); | ||
ImageProcessor imagePostProcessor = | ||
new ImageProcessor.Builder() | ||
.add(new DequantizeOp(/* zeroPoint= */ 0f, /* scale= */ 255f)) | ||
.build(); | ||
TensorImage outputTensorImage = new TensorImage(DataType.FLOAT32); | ||
outputTensorImage.load(outputImage); | ||
Log.w(TAG, "- Load output " + (System.currentTimeMillis() - before) + " ms"); | ||
|
||
before = System.currentTimeMillis(); | ||
Bitmap outputBitmap = imagePostProcessor.process(outputTensorImage).getBitmap(); | ||
Log.w(TAG, "- Post process output " + (System.currentTimeMillis() - before) + " ms"); | ||
|
||
texId = | ||
GlUtil.createTexture( | ||
outputBitmap.getWidth(), | ||
outputBitmap.getHeight(), | ||
/* useHighPrecisionColorComponents= */ false); | ||
GLES20.glBindTexture(GLES20.GL_TEXTURE_2D, texId); | ||
GLES20.glTexParameterf(GL10.GL_TEXTURE_2D, GL10.GL_TEXTURE_MIN_FILTER, GL10.GL_NEAREST); | ||
GLES20.glTexParameterf(GL10.GL_TEXTURE_2D, GL10.GL_TEXTURE_MAG_FILTER, GL10.GL_LINEAR); | ||
GLES20.glTexParameterf(GL10.GL_TEXTURE_2D, GL10.GL_TEXTURE_WRAP_S, GL10.GL_REPEAT); | ||
GLES20.glTexParameterf(GL10.GL_TEXTURE_2D, GL10.GL_TEXTURE_WRAP_T, GL10.GL_REPEAT); | ||
GLUtils.texImage2D(GLES20.GL_TEXTURE_2D, /* level= */ 0, outputBitmap, /* border= */ 0); | ||
GlUtil.checkGlError(); | ||
|
||
GlUtil.focusFramebufferUsingCurrentContext(boundFramebuffer[0], width, height); | ||
|
||
glProgram.use(); | ||
glProgram.setSamplerTexIdUniform("uTexSampler", texId, /* texUnitIndex= */ 0); | ||
float[] identityMatrix = GlUtil.create4x4IdentityMatrix(); | ||
glProgram.setFloatsUniform("uTexTransformationMatrix", identityMatrix); | ||
glProgram.setFloatsUniform("uTransformationMatrix", identityMatrix); | ||
glProgram.setBufferAttribute( | ||
"aFramePosition", | ||
GlUtil.getNormalizedCoordinateBounds(), | ||
GlUtil.HOMOGENEOUS_COORDINATE_VECTOR_SIZE); | ||
glProgram.bindAttributesAndUniforms(); | ||
|
||
GLES20.glDrawArrays(GLES20.GL_TRIANGLE_STRIP, /* first= */ 0, /* count= */ 4); | ||
GlUtil.checkGlError(); | ||
|
||
GlUtil.deleteTexture(texId); | ||
} catch (GlUtil.GlException e) { | ||
throw VideoFrameProcessingException.from(e); | ||
} | ||
} | ||
|
||
private static TensorImage getScaledTensorImage( | ||
Bitmap bitmap, int targetWidth, int targetHeight) { | ||
int cropSize = Math.min(bitmap.getWidth(), bitmap.getHeight()); | ||
ImageProcessor imageProcessor = | ||
new ImageProcessor.Builder() | ||
.add(new ResizeWithCropOrPadOp(cropSize, cropSize)) | ||
.add( | ||
new ResizeOp( | ||
targetHeight, | ||
targetWidth, | ||
ResizeOp.ResizeMethod.BILINEAR)) // TODO: Not sure why they are swapped? | ||
.add(new NormalizeOp(/* mean= */ 0f, /* stddev= */ 255f)) | ||
.build(); | ||
TensorImage tensorImage = new TensorImage(DataType.FLOAT32); | ||
tensorImage.load(bitmap); | ||
return imageProcessor.process(tensorImage); | ||
} | ||
} |
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Should we aim to convert it to Kotlin? Do you aim to keep this repo Kotlin only or any guidelines on this?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yes, we should convert this to Kotlin even though we don't have any strict guidelines on keeping this repo kotlin only.
so i do this in a separate PR to keep it easier to review or just do it as part of this PR?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I am ok with having a follow up PR with a Kotlin version as long as the repository owner is fine with it.