mirror of
https://github.com/mollyim/webrtc.git
synced 2025-05-14 14:20:45 +01:00

TestConfig: Replace Print method with ToString and add test. Bug: none Change-Id: I9853cb16875199a51c5731d1cec326159751d001 Reviewed-on: https://webrtc-review.googlesource.com/14320 Commit-Queue: Åsa Persson <asapersson@webrtc.org> Reviewed-by: Rasmus Brandt <brandtr@webrtc.org> Cr-Commit-Position: refs/heads/master@{#20420}
245 lines
9.3 KiB
C++
245 lines
9.3 KiB
C++
/*
|
|
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
|
|
*
|
|
* Use of this source code is governed by a BSD-style license
|
|
* that can be found in the LICENSE file in the root of the source
|
|
* tree. An additional intellectual property rights grant can be found
|
|
* in the file PATENTS. All contributing project authors may
|
|
* be found in the AUTHORS file in the root of the source tree.
|
|
*/
|
|
|
|
#ifndef MODULES_VIDEO_CODING_CODECS_TEST_VIDEOPROCESSOR_H_
|
|
#define MODULES_VIDEO_CODING_CODECS_TEST_VIDEOPROCESSOR_H_
|
|
|
|
#include <map>
|
|
#include <memory>
|
|
#include <string>
|
|
#include <vector>
|
|
|
|
#include "api/video/video_frame.h"
|
|
#include "common_video/libyuv/include/webrtc_libyuv.h"
|
|
#include "modules/video_coding/codecs/h264/include/h264_globals.h"
|
|
#include "modules/video_coding/codecs/test/packet_manipulator.h"
|
|
#include "modules/video_coding/codecs/test/stats.h"
|
|
#include "modules/video_coding/codecs/test/test_config.h"
|
|
#include "modules/video_coding/include/video_codec_interface.h"
|
|
#include "modules/video_coding/utility/ivf_file_writer.h"
|
|
#include "modules/video_coding/utility/vp8_header_parser.h"
|
|
#include "modules/video_coding/utility/vp9_uncompressed_header_parser.h"
|
|
#include "rtc_base/buffer.h"
|
|
#include "rtc_base/checks.h"
|
|
#include "rtc_base/constructormagic.h"
|
|
#include "rtc_base/sequenced_task_checker.h"
|
|
#include "rtc_base/task_queue.h"
|
|
#include "test/testsupport/frame_reader.h"
|
|
#include "test/testsupport/frame_writer.h"
|
|
|
|
namespace webrtc {
|
|
|
|
class VideoBitrateAllocator;
|
|
|
|
namespace test {
|
|
|
|
// Handles encoding/decoding of video using the VideoEncoder/VideoDecoder
|
|
// interfaces. This is done in a sequential manner in order to be able to
|
|
// measure times properly.
|
|
// The class processes a frame at the time for the configured input file.
|
|
// It maintains state of where in the source input file the processing is at.
|
|
//
|
|
// Regarding packet loss: Note that keyframes are excluded (first or all
|
|
// depending on the ExcludeFrameTypes setting). This is because if key frames
|
|
// would be altered, all the following delta frames would be pretty much
|
|
// worthless. VP8 has an error-resilience feature that makes it able to handle
|
|
// packet loss in key non-first keyframes, which is why only the first is
|
|
// excluded by default.
|
|
// Packet loss in such important frames is handled on a higher level in the
|
|
// Video Engine, where signaling would request a retransmit of the lost packets,
|
|
// since they're so important.
|
|
//
|
|
// Note this class is not thread safe and is meant for simple testing purposes.
|
|
class VideoProcessor {
|
|
public:
|
|
VideoProcessor(webrtc::VideoEncoder* encoder,
|
|
webrtc::VideoDecoder* decoder,
|
|
FrameReader* analysis_frame_reader,
|
|
FrameWriter* analysis_frame_writer,
|
|
PacketManipulator* packet_manipulator,
|
|
const TestConfig& config,
|
|
Stats* stats,
|
|
IvfFileWriter* encoded_frame_writer,
|
|
FrameWriter* decoded_frame_writer);
|
|
~VideoProcessor();
|
|
|
|
// Reads a frame from the analysis frame reader and sends it to the encoder.
|
|
// When the encode callback is received, the encoded frame is sent to the
|
|
// decoder. The decoded frame is written to disk by the analysis frame writer.
|
|
// Objective video quality metrics can thus be calculated after the fact.
|
|
void ProcessFrame();
|
|
|
|
// Updates the encoder with target rates. Must be called at least once.
|
|
void SetRates(int bitrate_kbps, int framerate_fps);
|
|
|
|
// Returns the number of dropped frames.
|
|
std::vector<int> NumberDroppedFramesPerRateUpdate() const;
|
|
|
|
// Returns the number of spatial resizes.
|
|
std::vector<int> NumberSpatialResizesPerRateUpdate() const;
|
|
|
|
private:
|
|
class VideoProcessorEncodeCompleteCallback
|
|
: public webrtc::EncodedImageCallback {
|
|
public:
|
|
explicit VideoProcessorEncodeCompleteCallback(
|
|
VideoProcessor* video_processor)
|
|
: video_processor_(video_processor),
|
|
task_queue_(rtc::TaskQueue::Current()) {}
|
|
|
|
Result OnEncodedImage(
|
|
const webrtc::EncodedImage& encoded_image,
|
|
const webrtc::CodecSpecificInfo* codec_specific_info,
|
|
const webrtc::RTPFragmentationHeader* fragmentation) override {
|
|
RTC_CHECK(codec_specific_info);
|
|
|
|
if (task_queue_ && !task_queue_->IsCurrent()) {
|
|
task_queue_->PostTask(
|
|
std::unique_ptr<rtc::QueuedTask>(new EncodeCallbackTask(
|
|
video_processor_, encoded_image, codec_specific_info)));
|
|
return Result(Result::OK, 0);
|
|
}
|
|
|
|
video_processor_->FrameEncoded(codec_specific_info->codecType,
|
|
encoded_image);
|
|
return Result(Result::OK, 0);
|
|
}
|
|
|
|
private:
|
|
class EncodeCallbackTask : public rtc::QueuedTask {
|
|
public:
|
|
EncodeCallbackTask(VideoProcessor* video_processor,
|
|
const webrtc::EncodedImage& encoded_image,
|
|
const webrtc::CodecSpecificInfo* codec_specific_info)
|
|
: video_processor_(video_processor),
|
|
buffer_(encoded_image._buffer, encoded_image._length),
|
|
encoded_image_(encoded_image),
|
|
codec_specific_info_(*codec_specific_info) {
|
|
encoded_image_._buffer = buffer_.data();
|
|
}
|
|
|
|
bool Run() override {
|
|
video_processor_->FrameEncoded(codec_specific_info_.codecType,
|
|
encoded_image_);
|
|
return true;
|
|
}
|
|
|
|
private:
|
|
VideoProcessor* const video_processor_;
|
|
rtc::Buffer buffer_;
|
|
webrtc::EncodedImage encoded_image_;
|
|
const webrtc::CodecSpecificInfo codec_specific_info_;
|
|
};
|
|
|
|
VideoProcessor* const video_processor_;
|
|
rtc::TaskQueue* const task_queue_;
|
|
};
|
|
|
|
class VideoProcessorDecodeCompleteCallback
|
|
: public webrtc::DecodedImageCallback {
|
|
public:
|
|
explicit VideoProcessorDecodeCompleteCallback(
|
|
VideoProcessor* video_processor)
|
|
: video_processor_(video_processor),
|
|
task_queue_(rtc::TaskQueue::Current()) {}
|
|
|
|
int32_t Decoded(webrtc::VideoFrame& image) override {
|
|
if (task_queue_ && !task_queue_->IsCurrent()) {
|
|
task_queue_->PostTask(
|
|
[this, image]() { video_processor_->FrameDecoded(image); });
|
|
return 0;
|
|
}
|
|
video_processor_->FrameDecoded(image);
|
|
return 0;
|
|
}
|
|
|
|
int32_t Decoded(webrtc::VideoFrame& image,
|
|
int64_t decode_time_ms) override {
|
|
return Decoded(image);
|
|
}
|
|
|
|
void Decoded(webrtc::VideoFrame& image,
|
|
rtc::Optional<int32_t> decode_time_ms,
|
|
rtc::Optional<uint8_t> qp) override {
|
|
Decoded(image);
|
|
}
|
|
|
|
private:
|
|
VideoProcessor* const video_processor_;
|
|
rtc::TaskQueue* const task_queue_;
|
|
};
|
|
|
|
// Invoked by the callback adapter when a frame has completed encoding.
|
|
void FrameEncoded(webrtc::VideoCodecType codec,
|
|
const webrtc::EncodedImage& encodedImage);
|
|
|
|
// Invoked by the callback adapter when a frame has completed decoding.
|
|
void FrameDecoded(const webrtc::VideoFrame& image);
|
|
|
|
void WriteDecodedFrameToFile(rtc::Buffer* buffer);
|
|
bool ExcludeFrame(const EncodedImage& encoded_image);
|
|
|
|
TestConfig config_ RTC_GUARDED_BY(sequence_checker_);
|
|
|
|
webrtc::VideoEncoder* const encoder_;
|
|
webrtc::VideoDecoder* const decoder_;
|
|
const std::unique_ptr<VideoBitrateAllocator> bitrate_allocator_;
|
|
|
|
// Adapters for the codec callbacks.
|
|
VideoProcessorEncodeCompleteCallback encode_callback_;
|
|
VideoProcessorDecodeCompleteCallback decode_callback_;
|
|
|
|
// Fake network.
|
|
PacketManipulator* const packet_manipulator_;
|
|
|
|
// These (mandatory) file manipulators are used for, e.g., objective PSNR and
|
|
// SSIM calculations at the end of a test run.
|
|
FrameReader* const analysis_frame_reader_;
|
|
FrameWriter* const analysis_frame_writer_;
|
|
|
|
// These (optional) file writers are used to persistently store the encoded
|
|
// and decoded bitstreams. The purpose is to give the experimenter an option
|
|
// to subjectively evaluate the quality of the processing. Each frame writer
|
|
// is enabled by being non-null.
|
|
IvfFileWriter* const encoded_frame_writer_;
|
|
FrameWriter* const decoded_frame_writer_;
|
|
|
|
// Keep track of inputed/encoded/decoded frames, so we can detect frame drops.
|
|
int last_inputed_frame_num_ RTC_GUARDED_BY(sequence_checker_);
|
|
int last_encoded_frame_num_ RTC_GUARDED_BY(sequence_checker_);
|
|
int last_decoded_frame_num_ RTC_GUARDED_BY(sequence_checker_);
|
|
|
|
// Store an RTP timestamp -> frame number map, since the timestamps are
|
|
// based off of the frame rate, which can change mid-test.
|
|
std::map<uint32_t, int> rtp_timestamp_to_frame_num_
|
|
RTC_GUARDED_BY(sequence_checker_);
|
|
|
|
// Keep track of if we have excluded the first key frame from packet loss.
|
|
bool first_key_frame_has_been_excluded_ RTC_GUARDED_BY(sequence_checker_);
|
|
|
|
// Keep track of the last successfully decoded frame, since we write that
|
|
// frame to disk when decoding fails.
|
|
rtc::Buffer last_decoded_frame_buffer_ RTC_GUARDED_BY(sequence_checker_);
|
|
|
|
// Statistics.
|
|
Stats* stats_;
|
|
std::vector<int> num_dropped_frames_ RTC_GUARDED_BY(sequence_checker_);
|
|
std::vector<int> num_spatial_resizes_ RTC_GUARDED_BY(sequence_checker_);
|
|
int rate_update_index_ RTC_GUARDED_BY(sequence_checker_);
|
|
|
|
rtc::SequencedTaskChecker sequence_checker_;
|
|
|
|
RTC_DISALLOW_COPY_AND_ASSIGN(VideoProcessor);
|
|
};
|
|
|
|
} // namespace test
|
|
} // namespace webrtc
|
|
|
|
#endif // MODULES_VIDEO_CODING_CODECS_TEST_VIDEOPROCESSOR_H_
|