Reduce a Caffe network model - c++

I'd like to use Caffe to extract image features. However, it takes too long to process an image, so I'm looking for ways to optimize for speed.
One thing I noticed is that the network definition I'm using has four extra layers on top the one from which I'm reading a result (and there are no feedback signals, so they should be safe to delete).
I tried to delete them from the definition file but it had no effect at all. I guess I might need to remove the corresponding part of the file that contains pre-trained weights, too. That is, however, a binary file (a protobuffer) so editing it is not that easy.
Do you think that removing the four layers might have a profound effect of the net performance?
If so then how do I get familiar with the file contents so that I could edit it and how do I know which parts to remove?

first, I don't think removing the binary weights will have any effect.
Second, you can do it easily using the python interface: see this tutorial.
Last but not least, have you tried running caffe time to measure the performance of your net? this may help you identify the bottlenecks of your computations.
PS,
You might find this thread relevant as well.

Caffemodel stores data as key-value pair. Caffe only copies weight for those layers (in train.prototxt) having exactly same name as caffemodel. Hence I don't think removing binary weights will work. If you want to change network structure, just modify train.prototxt and deploy.txt.
If you insist to remove weights from binary file, follow this caffe example.
And to make sure you delete right part, this visualizing tool should help.

I would retrain on a smaller input size, change strides, etc. However if you want to reduce file size, I'd suggest quantizing the weights https://github.com/yuanyuanli85/CaffeModelCompression and then using something like lzma compression (xz for unix). We do this so we can deploy to mobile devices. 8 bit weights compress nicely.

Related

Text Detection with YOLO on Challenging Images

I have images that look as follows:
My goal is to detect and recognize the number 31197394. I have already fine-tuned a deep neural network on text recognition. It can successfully identify the correct number, if it is provided it in the following format:
The only task that remains is the detection of the corresponding bounding box. For this purpose, I have tried darknet. Unfortunately, it's not recognizing anything. Does anyone have an idea of a network that performs better on these kind of images? I know, that amazon recognition is able to solve this task. But I need a solution that works offline. So my hopes are still high that there exist pre-trained networks that work. Thank's a lot for your help!
Don't say darknet doesn't work. It depends on how you labeled your dataset. It is true that the numbers you want to recognize are too small so if you don't make any changes to the image during the pre-processing phase, it would be complicated for a neural network to recognize them well. So what you can do that will surely work is:
1---> Before labeling, increase the size of all images by 2 times its current size (like 1000*1000)
2---> used this size (1000 * 1000) for the darket trainer instead of the default size proposed by darknet which is 416 * 416. You would then have to change the configuration file
3---> use the latest darknet version (yolo v4)
4---> On the configuration file, always keep a number of subdivisions at 1.
I also specify that this method is too greedy in memory, it is therefore necessary to provide a machine with RAM > 16 GB. The advantage is that it works...
Thanks for your answers guys! You were right, I had to finetune yolo to make it work. So I created a dataset and fine-tuned yolov5. I am surprised how good the results are. Despite only having about 300 images in total, I get an accuracy of 97% to predict the correct number. This is mainly due to strong augmentations. And indeed the memory requirements are large, but I could train on a 32 GM RAM machine. I can really encourage anyone who faces similar problems to give yolo a shot!!
Maybe use an R-CNN to identify the region where the number is and then pass that region to your fine-tuned neural network for the digit classification

Style transfer on large image. (in chunks?)

I am looking into various style transfer models and I noted that they all have limited resolution (when running on Pixel 3, for example, I couldn't go beyond 1,024x1,024, OOM otherwise).
I've noticed a few apps (eg this app) which appear to be doing style transfer for up to ~10MP images, these apps also show progress bar which I guess means that they don't just call a single tensorflow "run" method for entire image as otherwise they won't know how much was processed.
I would guess they are using some sort of tiling, but naively splitting the image into 256x256 produces inconsistent style (not just on the borders).
As this seems like an obvious problem I tried to find any publications about this, but I couldn't find any. Am I missing something?
Thanks!
I would guess people split the model into multiple ones (for VGG it is easy to do manually, eg. via layers) and then use model_summary Keras function (or benchmarks) to estimate relative time it takes for each step and thus guide progress bar. Such separation probably also saves memory as tensorflow lite might not be clever enough to reuse memory storing intermediate activations from lower layers once they are not needed.

Caffe Batch processing no speedup

I would like to speedup the forward pass of classification of a CNN using caffe.
I have tried batch classification in Caffe using code provided in here:
Modifying the Caffe C++ prediction code for multiple inputs
This solution enables me to give a vector of Mat, but it does not speed up anything. Even though the input layer is modified.
I am processing pretty small images (3x64x64) on a powerful pc with two GTX1080, and there is no issue in terms of memory.
I tried also changing the deploy.prototxt, but I get the same result.
It seems that at one point the forward pass of the CNN becomes sequential.
I have seen someone pointing this out here also:
Batch processing mode in Caffe - no performance gains
Another similar thread, for python : batch size does not work for caffe with deploy.prototxt
I have seen some things about MemoryDataLayer, but I am not sure this will solve my problem.
So I am kind of lost on what to do exactly... does anyone have any information on how to speedup classification time.
Thanks for any help !

Writing to the middle of the file (without overwriting data)

In windows is it possible through an API to write to the middle of a file without overwriting any data and without having to rewrite everything after that?
If it's possible then I believe it will obviously fragment the file; how many times can I do it before it becomes a serious problem?
If it's not possible what approach/workaround is usually taken? Re-writing everything after the insertion point becomes prohibitive really quickly with big (ie, gigabytes) files.
Note: I can't avoid having to write to the middle. Think of the application as a text editor for huge files where the user types stuff and then saves. I also can't split the files in several smaller ones.
I'm unaware of any way to do this if the interim result you need is a flat file that can be used by other applications other than the editor. If you want a flat file to be produced, you will have to update it from the change point to the end of file, since it's really just a sequential file.
But the italics are there for good reason. If you can control the file format, you have some options. Some versions of MS Word had a quick-save feature where they didn't rewrite the entire document, rather they appended a delta record to the end of the file. Then, when re-reading the file, it applied all the deltas in order so that what you ended up with was the right file. This obviously won't work if the saved file has to be usable immediately to another application that doesn't understand the file format.
What I'm proposing there is to not store the file as text. Use an intermediate form that you can efficiently edit and save, then have a step which converts that to a usable text file infrequently (e.g., on editor exit). That way, the user can save as much as they want but the time-expensive operation won't have as much of an impact.
Beyond that, there are some other possibilities.
Memory-mapping (rather than loading) the file may provide efficiences which would speed things up. You'd probably still have to rewrite to the end of the file but it would be happening at a lower level in the OS.
If the primary reason you want fast save is to start letting the user keep working (rather than having the file available to another application), you could farm the save operation out to a separate thread and return control to the user immediately. Then you would need synchronisation between the two threads to prevent the user modifying data yet to be saved to disk.
The realistic answer is no. Your only real choices are to rewrite from the point of the modification, or build a more complex format that uses something like an index to tell how to arrange records into their intended order.
From a purely theoretical viewpoint, you could sort of do it under just the right circumstances. Using FAT (for example, but most other file systems have at least some degree of similarity) you could go in and directly manipulate the FAT. The FAT is basically a linked list of clusters that make up a file. You could modify that linked list to add a new cluster in the middle of a file, and then write your new data to that cluster you added.
Please note that I said purely theoretical. Doing this kind of manipulation under a complete unprotected system like MS-DOS would have been difficult but bordering on reasonable. With most newer systems, doing the modification at all would generally be pretty difficult. Most modern file systems are also (considerably) more complex than FAT, which would add further difficulty to the implementation. In theory it's still possible -- in fact, it's now thoroughly insane to even contemplate, where it was once almost reasonable.
I'm not sure about the format of your file but you could make it 'record' based.
Write your data in chunks and give each chunk an id.
Id could be data offset in file.
At the start of the file you could
have a header with a list of ids so
that you can read records in
order.
At the end of 'list of ids' you could point to another location in the file (and id/offset) that stores another list of ids
Something similar to filesystem.
To add new data you append them at the end and update index (add id to the list).
You have to figure out how to handle delete record and update.
If records are of the same size then to delete you can just mark it empty and next time reuse it with appropriate updates to index table.
Probably the most efficient way to do this (if you really want to do it) is to call ReadFileScatter() to read the chunks before and after the insertion point, insert the new data in the middle of the FILE_SEGMENT_ELEMENT[3] list, and call WriteFileGather(). Yes, this involves moving bytes on disk. But you leave the hard parts to the OS.
If using .NET 4 try a memory-mapped file if you have an editor-like application - might jsut be the ticket. Something like this (I didn't type it into VS so not sure if I got the syntax right):
MemoryMappedFile bigFile = MemoryMappedFile.CreateFromFile(
new FileStream(#"C:\bigfile.dat", FileMode.Create),
"BigFileMemMapped",
1024 * 1024,
MemoryMappedFileAccess.ReadWrite);
MemoryMappedViewAccessor view = MemoryMapped.CreateViewAccessor();
int offset = 1000000000;
view.Write<ObjectType>(offset, ref MyObject);
I noted both paxdiablo's answer on dealing with other applications, and Matteo Italia's comment on Installable File Systems. That made me realize there's another non-trivial solution.
Using reparse points, you can create a "virtual" file from a base file plus deltas. Any application unaware of this method will see a continuous range of bytes, as the deltas are applied on the fly by a file system filter. For small deltas (total <16 KB), the delta information can be stored in the reparse point itself; larger deltas can be placed in an alternative data stream. Non-trivial of course.
I know that this question is marked "Windows", but I'll still add my $0.05 and say that on Linux it is possible to both insert or remove a lump of data to/from the middle of a file without either leaving a hole or copying the second half forward/backward:
fallocate(fd, FALLOC_FL_COLLAPSE_RANGE, offset, len)
fallocate(fd, FALLOC_FL_INSERT_RANGE, offset, len)
Again, I know that this probably won't help the OP but I personally landed here searching for a Linix-specific answer. (There is no "Windows" word in the question, so web search engine saw no problem with sending me here.

How do I write a Perl script to filter out digital pictures that have been doctored?

Last night before going to bed, I browsed through the Scalar Data section of Learning Perl again and came across the following sentence:
the ability to have any character in a string means you can create, scan, and manipulate raw binary data as strings.
An idea immediately hit me that I could actually let Perl scan the pictures that I have stored on my hard disk to check if they contain the string Adobe. It seems by doing so, I can tell which of them have been photoshopped. So I tried to implement the idea and came up with the following code:
#!perl
use autodie;
use strict;
use warnings;
{
local $/="\n\n";
my $dir = 'f:/TestPix/';
my #pix = glob "$dir/*";
foreach my $file (#pix) {
open my $pic,'<', "$file";
while(<$pic>) {
if (/Adobe/) {
print "$file\n";
}
}
}
}
Excitingly, the code seems to be really working and it does the job of filtering out the pictures that have been photoshopped. But problem is many pictures are edited by other utilities. I think I'm kind of stuck there. Do we have some simple but universal method to tell if a digital picture has been edited or not, something like
if (!= /the origianl format/) {...}
Or do we simply have to add more conditions? like
if (/Adobe/|/ACDSee/|/some other picture editors/)
Any ideas on this? Or am I oversimplifying due to my miserably limited programming knowledge?
Thanks, as always, for any guidance.
Your best bet in Perl is probably ExifTool. This gives you access to whatever non-image information is embedded into the image. However, as other people said, it's possible to strip this information out, of course.
I'm not going to say there is absolutely no way to detect alterations in an image, but the problem is extremely difficult.
The only person I know of who claims to have an answer is Dr. Neal Krawetz, who claims that digitally altered parts of an image will have different compression error rates from the original portions. He claims that re-saving a JPEG at different quality levels will highlight these differences.
I have not found this to be the case, in my investigations, but perhaps you might have better results.
No. There is no functional distinction between a perfectly edited image, and one which was the way it is from the start - it's all just a bag of pixels in the end, after all, and any other metadata you can remove or forge all you want.
The name of the graphics program used to edit the image is not part of the image data itself but of something called meta data - which may be stored in the image file but, as others have noted, is neither required (so some programs may not store it, some may allow you an option of not storing it) nor reliable - if you forged an image, you might have forged the meta data as well.
So the answer to your question is "no, there's no way to universally tell if the pic was edited or not, although some image editing software may write its signature into the image file and it'll be left there by carelessness of the editing person.
If you're inclined to learn more about image processing in Perl, you could take a look at some of the excellent modules CPAN has to offer:
Image::Magick - read, manipulate and write of a large number of image file formats
GD - create colour drawings using a large number of graphics primitives, and emit the drawings in various formats.
GD::Graph - create charts
GD::Graph3d - create 3D Graphs with GD and GD::Graph
However, there are other utilities available for identifying various image formats. It's more of a question for Super User, but for various unix distros you can use file to identify many different types of files, and for MacOSX, Graphic Converter has never let me down. (It was even able to open the bizarre multi-file X-ray of my cat's shattered pelvis that I got on a disc from the vet.)
How would you know what the original format was? I'm pretty sure there's no guaranteed way to tell if an image has been modified.
I can just open the file (with my favourite programming language and filesystem API) and just write whatever I want into that file willy-nilly. As long as I don't screw something up with the file format, you'd never know it happened.
Heck, I could print the image out and then scan it back in; how would you tell it from an original?
As other's have stated, there is no way to know if the image was doctored. I'm guessing what you basically want to know is the difference between a realistic photograph and one that has been enhanced or modified.
There's always the option of running some extremely complex image recognition algorithm that would analyze every pixel in your image and do some very complicated stuff to determine if the image was doctored or not. This solution would probably involve AI which would examine millions of photos that are both doctored and those that are not and learn from them. However, this is more of a theoretical solution and isn't very practical... you would probably only see it in movies. It would be extremely complex to develop and probably take years. And even if you did get something like this to work, it probably still wouldn't be 100% correct all the time. I'm guessing AI technology still isn't at that level and could take a while until it is.
A not-commonly-known feature of exiftool allows you to recognize the originating software through an analysis of the JPEG quantization tables (not relying on image metadata). It recognizes tables written by many applications. Note that some cameras may use the same quantization tables as some applications, so this isn't a 100% solution, but it is worth looking into. Here is an example of exiftool run on two images, the first was edited by photoshop.
> exiftool -jpegdigest a.jpg b.jpg
======== a.jpg
JPEG Digest : Adobe Photoshop, Quality 10
======== b.jpg
JPEG Digest : Canon EOS 30D/40D/50D/300D, Normal
2 image files read
This will work even if the metadata has been removed.
There is existing software out there which uses various techniques (compression artifacting, comparison to signature profiles in a database of cameras, etc.) to analyze the actual image data for evidence of alteration. If you have access to such software and the software available to you provides an API for external access to these analysis functions, then there's a decent chance that a Perl module exists which will interface with that API and, if no such module exists, it could probably be created rather quickly.
In theory, it would also be possible to implement the image analysis code directly in native Perl, but I'm not aware of anyone having done so and I expect that you'd be better off writing something that low-level and processor-intensive in a fully-compiled language (e.g., C/C++) rather than in Perl.
http://www.impulseadventure.com/photo/jpeg-snoop.html
is a tool that does the job almost good
If there has been any cloning , there is a variation in the pixel density..or concentration which sometimes shows up.. upon manual inspection
a Photoshop cloned area will have even pixel density(my meaning is variation of Pixels wrt a scanned image)