I'm using mp4parser to mux h264 and aac file which are re-encoded from orginal video file,how can I write the metadata of the original video to the new mp4 file? Or is there a common method to write metadata to mp4 file?
metadata and MP4 is a really problem. There is no generally supported specification. But this is only one part of the problem.
Prob (1): When to write metadata
Prob (2): What to write
Prob (1) is relatively easy to solve: Just extend the DefaultMp4Builder or the FragmentedMp4Builder on your own and override the
protected ParsableBox createUdta(Movie movie) {
return null;
}
with something meaningful. E.g.:
protected ParsableBox createUdta(Movie movie) {
UserDataBox udta = new UserDataBox();
CopyrightBox copyrightBox = new CopyrightBox();
copyrightBox.setCopyright("All Rights Reserved, me, myself and I, 2015");
copyrightBox.setLanguage("eng");
udta.addBox(copyrightBox);
return udta;
}
some people used that to write apple compatible metadata but even though there are some classes in my code I never really figured out what works and what not. You might want to have a look into Apple's specification here
And yes: I'm posting this a year to late.
It seems that the 'mp4parser' library (https://code.google.com/p/mp4parser/), supports writing Metadata to mp4 files in Android. However, I've found there's little-to-no documentation on how to do this, beyond a few examples in their codebase. I've had some luck with the following example, which writes XML metadata into the 'moov/udta/meta' box:
https://github.com/copiousfreetime/mp4parser/blob/master/examples/src/main/java/com/googlecode/mp4parser/stuff/ChangeMetaData.java
If you consider the alternatives you might want to look at JCodec for this purpose. It now has the org.jcodec.movtool.MetadataEditor API (and a matching CLI org.jcodec.movtool.MetadataEditorMain).
Their documentation contains many samples: http://jcodec.org/docs/working_with_mp4_metadata.html
So basically when you want to add some metadata you need to know what key(s) it corresponds to. One way to find out is to inspect a sample file that already has the metadata you need. For this you can run the JCodec's CLI tool that will just print out all the existing metadata fields (keys with values):
./metaedit <file.mp4>
Then when you know the key you want to work with you can either use the same CLI tool:
# Changes the author of the movie
./metaedit -f -si ©ART=New\ value file.mov
or the same thing via the Java API:
MetadataEditor mediaMeta = MetadataEditor.createFrom(new
File("file.mp4"));
Map<Integer, MetaValue> meta = mediaMeta.getItunesMeta();
meta.put(0xa9415254, MetaValue.createString("New value")); // fourcc for '©ART'
mediaMeta.save(false); // fast mode is off
To delete a metadata field from a file:
MetadataEditor mediaMeta = MetadataEditor.createFrom(new
File("file.mp4"));
Map<Integer, MetaValue> meta = mediaMeta.getItunesMeta();
meta.remove(0xa9415254); // removes the '©ART'
mediaMeta.save(false); // fast mode is off
To convert string to integer fourcc you can use something like:
byte[] bytes = "©ART".getBytes("iso8859-1");
int fourcc =
ByteBuffer.wrap(bytes).order(ByteOrder.BIG_ENDIAN).getInt();
If you want to edit/delete the android metadata you'll need to use a different set of fucntion (because it's stored differently than iTunes metadata):
./metaedit -sk com.android.capture.fps,float=25.0 file.mp4
OR alternatively the same through the API:
MetadataEditor mediaMeta = MetadataEditor.createFrom(new
File("file.mp4"));
Map<String, MetaValue> meta = mediaMeta.getKeyedMeta();
meta.put("com.android.capture.fps", MetaValue.createFloat(25.));
mediaMeta.save(false); // fast mode is off
Related
I tried a while to get the pretrained model working on android. The problem is, I only got the ckpt and meta file for the pretrained net. In my opinion I need the .pb for the android app. So I tried to convert the given files to an .pb file.
Therefore I tried the freeze_graph.py but without succes. So I used the example code from https://github.com/openimages/dataset/blob/master/tools/classify.py and modified it to store a pb. file after loading
if not os.path.exists(FLAGS.checkpoint):
tf.logging.fatal(
'Checkpoint %s does not exist. Have you download it? See tools/download_data.sh',
FLAGS.checkpoint)
g = tf.Graph()
with g.as_default():
input_image = tf.placeholder(tf.string)
processed_image = PreprocessImage(input_image)
with slim.arg_scope(inception.inception_v3_arg_scope()):
logits, end_points = inception.inception_v3(
processed_image, num_classes=FLAGS.num_classes, is_training=False)
predictions = end_points['multi_predictions'] = tf.nn.sigmoid(
logits, name='multi_predictions')
init_op = control_flow_ops.group(tf.global_variables_initializer(),
tf.global_variables_initializer(),
data_flow_ops.initialize_all_tables())
saver = tf_saver.Saver()
sess = tf.Session()
saver.restore(sess, FLAGS.checkpoint)
outpt_filename = 'output_graph.pb'
#output_graph_def = sess.graph.as_graph_def()
output_graph_def = graph_util.convert_variables_to_constants(sess, sess.graph.as_graph_def(), ["multi_predictions"])
with gfile.FastGFile(outpt_filename, 'wb') as f:
f.write(output_graph_def.SerializeToString())
Now my problem is that I have the .pb file but I don't have any opinion what is the input node name and I am not sure if multi_predictions is the right output name. In the example android app I have to specify both. And the android app crashed with:
tensorflow_inference_jni.cc:138 Could not create Tensorflow Graph: Invalid argument: No OpKernel was registered to support Op 'DecodeJpeg' with these attrs.
I don't know if there are more problem by trying to fix the .pb problem. Or if anyone knows a better way to port the ckpt and meta files to a .pd file in my case or knows a source for the final file with input and ouput names please give me a hint to complete this task.
Thanks
You'll need to use the optimize_for_inference.py script to strip out the unused nodes in your graph. "decodeJpeg" is not supported on Android -- pixel values should be fed in directly. ClassifierActivity.java has more detail about the specific nodes to use for inception v3.
In OpenCV I am using ORB FeatureDetection in an Android app. It has parameters but these cannot be set directly in Java. The recommended way to set them is to write out a XML or YML file with the params and then read it back in.
However, there seems to be no way to be sure that in fact I have written the file correctly and that the parameters are applied. I thought I could use the write() method to verify that my new settings have taken. But this doesn't seem to work. It does indeed write a proper YML or XML file, but the files have no params.
Here is code to write the files:
_detector = FeatureDetector.create(FeatureDetector.ORB);
// write initial params.
String fileName = myDir.getPath() + "/orb_params.yml";
_detector.write(fileName);
fileName = myDir.getPath() + "/orb_params.xml";
_detector.write(fileName);
// try setting some params.
String tempFileName = writeToFile("tempFile", "%YAML:1.0\nscaleFactor: 1.1\nnLevels: 5\nfirstLevel: 0\nedgeThreshold: 31\npatchSize: 31\n");
_detector.read(tempFileName);
// write params again.
String fileName = myDir.getPath() + "/orb_params2.yml";
_detector.write(fileName);
fileName = myDir.getPath() + "/orb_params2.xml";
_detector.write(fileName);
The xml files look like this:
<?xml version="1.0"?>
<opencv_storage>
</opencv_storage>
The yml files look like this:
%YAML:1.0
Is the write method not implemented in Java? I see that it is implemented in Algorithm.cpp and I assume the ORB Feature Detector just uses that implementation, I see no code to indicate otherwise.
I don't know if this problem is limited to ORB Feature Detection or if other algos suffer these problems when trying to write parameters from Java.
addendum: I see that the write method appears to be implemented in the JNI code: https://github.com/Itseez/opencv/blob/ddf82d0b154873510802ef75c53e628cd7b2cb13/modules/features2d/misc/java/src/cpp/features2d_manual.hpp
Got a response on answers.opencv.org and am posting it here for those stumbling across this in the future:
According to this answer the read/write function is not implemented in C++ and thus won't work in Android either.
I am working on a project in Flash Mobile using ActionScript. I have a zipped wav file that I need to be able to de serialize and play as needed in a Button Press action. Below is the code for zipping the wav file.
mic.removeEventListener(SampleDataEvent.SAMPLE_DATA, micSampleDataHandler);
btnRecord.setStyle("icon", recOff);
sampleCount++;
// save the raw PCM samples as a bare WAV file
var wav:ByteArray = new ByteArray();
var writer:WAVWriter = new WAVWriter();
writer.numOfChannels = 1;
writer.sampleBitRate = 16;
writer.samplingRate = 11025;
samples.position = 0;
writer.processSamples(wav, samples, 11025, 1);
wav.position = 0;
// zip the WAV file
var fzip:FZip = new FZip();
fzip.addFile(name + sampleCount.toString(), wav);
var zip:ByteArray = new ByteArray();
fzip.serialize(zip);
var recSpot:Object = {
id: null,
audio: zip,
name: "New Audio File " + newRecNum,
existsdb: "false"
};
newRecNum++;
recordings.addItem(recSpot);
}
What can I do to play this file, really haven't had to play a zipped file before.
I'm not familiar with WAVWriter (which is probably somewhat beside the point), but here's what I do know.
Firstly, because of the nature of a compression, you cannot (as far as I know) play a zipped audio file, period. You will need to unzip it first.
A quick Google search turned up THIS AS3 TUTORIAL on unzipping with FZIP. The example program is using .PNGs, but I would assume you can adjust it to work with the raw .WAV file you zipped earlier. Skip down to Step 5 for the actual code. (You'll need to rewrite it to work with your interface, obviously.)
You won't need the DataProvider variable in step 5, as that is for components, specifically. You'll need to load your data into something else. If your method of playing WAV files is anything like mine (I use the as3WAVSound class), you'll probably want to load the data into a ByteArray and play off of that.
You also probably won't need the for loop he uses in step 10, as your code appears to be creating a ZIP with only one WAV file. That simplifies things considerably.
Anyway, I hope that answers your question!
I have yet another hurdle to climb with my GOOGLE DRIVE SDK Android App. I am uploading scanned images with tightly controlled index fields - user defined 'tags' from local dictionary. For instance XXX.JPG has index words "car" + "insurance". Here is a simplified code snippet:
...
body.setTitle("XXX.JPG");
body.setDescription("car, insurance");
body.setIndexableText(new IndexableText().setText("car insurance"));
body.setMimeType("image/jpeg");
body.setParents(Arrays.asList(new ParentReference().setId(...)));
FileContent cont = new FileContent("image/jpeg", new java.io.File(fullPath("xxx.jpg")));
File gooFl = _svc.files().insert(body, cont).execute();
...
Again, everything works great, except when I start a search, I get results that apparently come from some OCR post process, thus rendering my system's DICTIONARY unusable. I assume I can use a custom MIME type, but then the JPEG images become invisible for users who use standard GOOGLE DRIVE application (local, browser-based ... ). So the question is: Can I upload MIME "image/jpeg" files with custom indexes (either Indexable, or Description fields) but stop GOOGLE from OCR-ing my files and adding indexes I did not intend to have?
Just to be more specific, I search for "car insurance" and instead of my 3 files I indexed this way, I get unmanageable pile of other results (JPEG scanned documents) that had "car" and "insurance" somewhere in them. Not what my app wants.
Thank you in advance, sean
...
Based on Burcu's advise below, I modified my code to something that looks like this (stripped to bare bones):
// define meta-data
File body = new File();
body.setTitle("xxx.jpg");
body.setDescription(tags);
body.setIndexableText(new IndexableText().setText(tags));
body.setMimeType("image/jpeg");
body.setParents(Arrays.asList(new ParentReference().setId(_ymID)));
body.setModifiedDate(DateTime.parseRfc3339(ymdGOO));
FileContent cont =
new FileContent("image/jpeg",new java.io.File(fullPath("xxx.jpg")));
String sID = findOnGOO(driveSvc, body.getTitle());
// file not found on gooDrive, upload and fix the date
if (sID == null) {
driveSvc.files().insert(body, cont).setOcr(false).execute();
driveSvc.files().patch(gooFl.getId(), body).setOcr(false).setSetModifiedDate(true).execute();
// file found on gooDrive - modify metadata and/or body
} else {
// modify content + metadata
if (contentModified) {
driveSvc.files().update(sID, body, cont).setOcr(false).setSetModifiedDate(true).execute();
// only metadata (tags,...)
} else {
driveSvc.files().patch(sID, body).setOcr(false).setSetModifiedDate(true).execute();
}
}
...
It is a block that uploads or modifies a Google Drive file. The two non-standard operations are:
1/ resetting the file's 'modified' date in order to force the date of file creation - tested, works OK
2/ stopping the OCR process that interferes with my apps indexing scheme - will test shortly and update here
For the sake of simplicity, I did not include the implementation of "findInGOO()" method. It is quite simple 2-liner and I can supply it upon request
sean
On insertion, set the ocr parameter to false:
service.files().update(body, content).setOcr(false).execute();
I have a set of mp4 files that I need to load onto an SD Card and read in my Android app.
The app needs to be able to search the videos by category, so my plan was to add some category info in the mp4 metadata before loading them (probably in the "description" field) and then use a ManagedQuery on the MediaStore.Video.Media.EXTERNAL_CONTENT_URI to find them.
I have updated the "description" field using Adobe Bridge, but when I look at the tags returned by a search, the "description" field is always null. Clearly, the data I'm writing to the mp4 files is not being picked up when Android looks at the video file.
Is there another way I should be writing/searching video metadata?
You should probably look at MediaStore.Video.VideoColumns.DESCRIPTION.
You can query the MediaStore.Video.Media.EXTERNAL_CONTENT_URI as you would any other content provider: http://developer.android.com/guide/topics/providers/content-providers.html#querying
You can add/delete/lookup metadata fields in MP4 files using JCodec's MetadataEditor class.
If you want to add metadata field to MP4 file you can use:
MetadataEditor mediaMeta = MetadataEditor.createFrom(new
File("file.mp4"));
Map<String, MetaValue> meta = mediaMeta.getKeyedMeta();
meta.put("com.android.capture.fps", MetaValue.createFloat(25.));
mediaMeta.save(false); // fast mode is off
Alternatively the same can be done from the command line with the CLI tool (MetadataEditorMain):
./metaedit -sk com.android.capture.fps,float=25.0 file.mp4
In the Java code you can get the list of metadata like so:
MetadataEditor mediaMeta = MetadataEditor.createFrom(new
File("file.mp4"));
Map<String, MetaValue> meta = mediaMeta.getKeyedMeta();
for (String key : meta.keySet()) {
System.out.println(key + ": " + meta.get(key));
}
OR from the command line:
./metaedit <file.mp4>
Read more about it: http://jcodec.org/docs/working_with_mp4_metadata.html