The tool runs the analysis between every two corresponding frames until either of the videos runs out of frames.
The results are written in a file (results_file provided from the command line) in the format:
Frame: <frame_number>, PSNR: <psnr_value>, SSIM: <ssim_value>
TEST=
psnr_ssim_analyzer --reference_file=<name_of_file>
--test_file=<name_of_file> --results_file=<name_of_file>
--width=<width_of_frames> --height=<height_of_frames>
Review URL: https://webrtc-codereview.appspot.com/748007
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2665 4adac7df-926f-26a2-2b94-8c16560cd09d
The bug was introduced in the last CL during optimization.
With the new inine ARM assembly, the resulted assembly code of the loop is 28 instructions, compared to 47 lines by the original.
The optimized (with WEBRTC_ARCH_ARM_V7A defined in This file) and un-optimized code (with WEBRTC_ARCH_ARM_V7A not defined in This file) are bit-exact.
Review URL: https://webrtc-codereview.appspot.com/748004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2651 4adac7df-926f-26a2-2b94-8c16560cd09d
- Imported the floating point implementation of iSAC coder from stable public webRTC release:
http://code.google.com/p/webrtc/source/browse/#svn%2Fstable
- Added a reminder (in a TODO file) to switch from floating point to fixed point implementation.
Please note that this is the minimal amount of functionality we (the speech text-to-speech team) require
for delivering our unit selection algorithms for the Glass project. I'd be happy to address any comments
you have.
Change-Id: I20481680786ec5372258efcea69ea5c9750a321a
This is a very simple tool which takes a test and reference YUV files and compares the frames in them.
The interesting part is that the test video is the video formed by the frames captured from the browser
(WebRTC output). The test video frames has been decoded and the correspondence between evrey frame in the
test video and the decoded barcode, i.e. the frame number in the refrence video has been written to a stats
file in the form frame_xxxx yyyy, where xxxx is the number of the frame in the test video and yyyy is the
number of the frame in the reference video.
We can have jumping over frames or duplicate frames in the test video, as well as incorrectly decoded barcodes.
The tool takes care of these cases.
I haven't used the video_metrics.h because the functions in there seem to do much more than we need and not to
do things that I needed.
The tool may need to be changed so that it could produce output which in turn will be used by PythonCharts or
other chart-drawing tool or library that we decide to use.
BUG=
TEST=
./out/Debug/frame_analyzer --reference_file=reference.yuv --test_file=test.yuv --stats_file=stats.txt --width=352 --height=288
Review URL: https://webrtc-codereview.appspot.com/701005
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2620 4adac7df-926f-26a2-2b94-8c16560cd09d
The barcode encoder generates barcodes as PNG files, than converts them to YUV files,
then stitches these into one YUV video file and finaly overlays this video over the
base video to produce a YUV video of the base with encoded barcodes in the upper part
of the video.
The decoder gets a YUV video with overlaid barcodes and first splits it into PNG files,
than decodes every file, and finally generates a stats file.
The encoder and decoder use Zxing and its Java command-line tools for the encoding and
decoding. They also use ffmpeg for the conversion between PNG and YUV and vice versa.
BUG=
TEST=
The tools could be used from trunk as:
./tools/barcode_tools/barcode_encoder.py --barcode_height=32 --base_frame_width=352
--base_frame_height=288 --base_yuv=<path_nad_name_to_base_file>
--output_yuv=<path_nad_name_to_output_file>
./tools/barcode_tools/barcode_decoder.py --yuv_file=<path_and_name_to_yuv_file>
--yuv_frame_width=352 --yuv=frame_height=288 --barcode_height=32
--stats_file=<path_and_name_of_stats_file>
By default the width of the barcodes is the same as the width of the base YUV frames.
This could be changes with the command-line option --barcode_width
Review URL: https://webrtc-codereview.appspot.com/679010
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2612 4adac7df-926f-26a2-2b94-8c16560cd09d
- Made the output file to open in write mode instead of append mode.
- Now the tool deletes the RGBA frames after conversion.
- Other minor cleanup work.
BUG=
TEST=
rgba_to_i420_converter --frames_dir=<directory_to_rgba_frames> --output_file=<output_yuv_file> --width=<width_of_input_frames> --height=<height_of_input_frames>
Review URL: https://webrtc-codereview.appspot.com/728004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2611 4adac7df-926f-26a2-2b94-8c16560cd09d
