thumbor / thumbor Goto Github PK

Create a graphics magick engine for thumbor.

Reference:
http://en.wikipedia.org/wiki/GraphicsMagick

Allow users to use a fit-in flag that specifies that the generated image MUST fit in the given width and height but keep the same proportion as the original.

Theres a sample code here:

http://love-python.blogspot.com/2008/10/get-remote-file-size-through-http.html

We could add an option that would limit the size of the downloaded image from the http loader.

We need to write a console app that gets installed alongside thumbor that allows users to call:

thumbor-encrypt --security-key="my-security-key" --width=300 --height=300 --smart /path/to/image.jpg

that outputs the encrypted url to help users easily obtain the url where toolkits are not available.

Create an auto-contrast filter, like:

/filter('auto-contrast')

AllowedSource has to be:

• Optional

Format has to change to be:

• Wildcarded - .mydomain.com
  (".globo.com", "*.glbimg.com")
• Regular Expression
  (r"(.+)[.]globo[.]com", r"(.+)[.]glbimg[.]com"

Currently thumbor's http_loader can only download one file at a time because it uses tornado's blocking http client. This causes image downloads to take a lot longer than they should.

You should pass a callback to the loader which will be called to process the image once the download is complete.

There isn't much point to your using tornado currently since you are blocking when loading images.

Thumbor is currently throwing errors while processing an image with its extension uppercased

Store in the storage the detection results for later usage.

Add context processors to thumbor that add in the request before thumbor processes it.

NoStorage Storage needs to be updated to include no crypto support

Allow users to pass a watermark to be added to the center of the image.

/filter('watermark', 'my.domain.com/watermark.png')

Allow specification of filters in configuration like:

Filters = [
"thumbor.filters.dropshadow",
"thumbor.filters.blur",
#etc
]

Filters should be added to the url like:

/300x200/smart/filter('blur', 0.3, 'gaussian')/my/image.jpg

The formation of filters is:

/filter('key', argument1, argument2, ...)

We can add a very simple filter in this story to test the infrastructure like the blur filter above.

Create an auto-color filter, like:

/filter('auto-color')

Drop a shadow "behind" the image.

/filter('shadow', 45, '#333', 0.8)

Arguments are angle, color and alpha percentage.

Trying to run thumbor-url with the latest version off of github.

Stacktrace:

thumbor thumbor-url http://google.com/cheese.jpg
Traceback (most recent call last):
File "/usr/local/topsoil_virtualenv/bin/thumbor-url", line 8, in
load_entry_point('thumbor==2.3.0', 'console_scripts', 'thumbor-url')()
File "/usr/local/topsoil_virtualenv/lib/python2.7/site-packages/thumbor/url_composer.py", line 50, in main
conf_file = ThumborServiceApp.get_conf_file('')
AttributeError: type object 'ThumborServiceApp' has no attribute 'get_conf_file'

Maybe something to do with latest version of tornado?

If an option called META_CALLBACK_NAME is specified in thumbor.conf then meta will always return as jsonp like:

callback({
#something
});

where "callback" is the value of META_CALLBACK_NAME.

Create a rounded corners filter, like:

/filter('rounded', 3, 3, 3, 3)

With the four arguments being the top-left, top-right, bottom-left and bottom-right corner radius in pixels.

Simply because pythonmagickwand causes great pain and suffering to install and work with.
The other engines work great and pythonmagickwand doesn't provide any great advantage that would justify going through the hassle of maintaining it.

Create a MySQL storage.

Create a python toolkit to expose the encrypt and decrypt functions of the crypto module in thumbor.

Looks like opencv just detect goodfeatures and faces on jpeg images. Thumbor should convert the image to jpeg, if it's not jpeg, and then try to detect features on it.

Allow users to specify a security key file on thumbor console,

Stop using tornado's options since they work at the thread level. Start using app-level options, since the App object is available to all handlers.

We have an issue that OpenCV does not deal properly with PIL's images, so it would be worth it if there was an actual file in the filesystem for OpenCV to Load.

This could be implemented with Python's namedtempfile.

Create an auto-tone filter, like:

/filter('auto-tone')

Using graphics magick and the URL http://localhost:8888/unsafe/fit-in/400x300/http://s.glbimg.com/jo/eg/f/original/2011/11/04/1________.jpg

Instead of getting an image with 400px of width, we get one with 399px of width.

Some sort of approximation error.

Allow users to modify the urls in thumbor both for Safe and Unsafe end-points.

Redis configuration in the config file does not work properly.

The face detection supplementary XML files are not installed when installing using pip or easy_install. Thumbor crashes saying OpenCV returned a null value.

Traceback (most recent call last):
File "/path/to/thumbor/virtualenv/lib/python2.6/site-packages/tornado/web.py", line 861, in _execute
getattr(self, self.request.method.lower())(_args, *_kwargs)
File "/path/to/thumbor/virtualenv/lib/python2.6/site-packages/tornado/web.py", line 941, in wrapper
return method(self, _args, *_kwargs)
File "/path/to/thumbor/virtualenv/lib/python2.6/site-packages/thumbor/handlers.py", line 268, in get
return self.execute_image_operations(opt, image)
File "/path/to/thumbor/virtualenv/lib/python2.6/site-packages/thumbor/handlers.py", line 74, in execute_image_operations
opt['smart'], image)
File "/path/to/thumbor/virtualenv/lib/python2.6/site-packages/thumbor/handlers.py", line 143, in get_image
self._fetch(image, extension, callback)
File "/path/to/thumbor/virtualenv/lib/python2.6/site-packages/thumbor/handlers.py", line 171, in _fetch
callback(buffer)
File "/path/to/thumbor/virtualenv/lib/python2.6/site-packages/thumbor/handlers.py", line 131, in callback
self.detectors[0](index=0, detectors=self.detectors).detect(context)
File "/path/to/thumbor/virtualenv/lib/python2.6/site-packages/thumbor/detectors/face_detector/init.py", line 26, in init
self.load_cascade_file(file, options.FACE_DETECTOR_CASCADE_FILE)
File "/path/to/thumbor/virtualenv/lib/python2.6/site-packages/thumbor/detectors/init.py", line 43, in load_cascade_file
setattr(self.class, 'cascade', cv.Load(cascade_file))
TypeError: OpenCV returned NULL

Unlike display the full URL address, display a name that the thumbor service recognizes (optional):
http:///unsafe/300x200/smart/glb1/et/bb/f/original/2011/03/24/VN0JiwzmOw0b0lg.jpg

http:///unsafe/300x200/smart/glb2/et/bb/f/original/2011/03/24/VN0JiwzmOw0b0lg.jpg

glb1 => s.glbimg.com
glb2 => s2.glbimg.com

Thanks

Cropping is not working as intended.

Create the OpenCV filter that allows users to use facial profile-face detection.

Create a NoStorage class that always returns False in the exists method and does nothing in the save method.

This is a storage for debugging or benchmarking purposes and should not be used in prod.

Support cryptography using a salt in the form of:

/cryptographed_url/some/nice/image.jpg

Then cryptographed url corresponds to:

/300x300/smart/image_path_hash/some/nice/image.jpg

The image path hash bit is to ensure that the original image path that was used for this size was for this url (to prevent tampering with the url).

Allow users to configure thumbor to use unlimited dimensions of images when configure MAX_WIDTH and MAX_HEIGHT to 0.

OpenCV code to find Red Eyes: http://www.betasix.net/programming/opencv/opencv-2-2-python-examples/

Hi Fabio,

I'm new to python and have some trouble get your thumbor running on my ubuntu regarding the Crypto.Cipher lib

I run setup.py, installed pycrypto and started app.py , but got an exception:

File "app.py", line 21, in
from thumbor.handlers.crypto import CryptoHandler
File "/usr/local/lib/python2.6/dist-packages/thumbor-0.8.3-py2.6.egg/thumbor/handlers/crypto.py", line 15, in
from thumbor.crypto import Crypto
File "/usr/local/lib/python2.6/dist-packages/thumbor-0.8.3-py2.6.egg/thumbor/crypto.py", line 14, in
from Crypto.Cipher import AES
ImportError: cannot import name AES

I even placed the Crypto source in thumbor/Cryptor and run a test app:

from Crypto.Cipher import DES3
from Crypto.Cipher import AES

In that case I did not get an error.

Fabio, do you know what I'm doing wrong?

cheers
seb

Meta endpoint should return what the image final dimensions are.

Change Mongo Storage to use GridFS.

This debug flag would draw rectangles for the faces detected and a point on the calculated center of mass on the image, so we can see easily if the face detection is working fine.

Supporting mongoDb as an image storage is pretty easy with MongoEngine(http://mongoengine.org/docs/v0.4/guide/gridfs.html) support to GridFS.

There should be a native MongoDB storage system in thumbor.

Not sure if this is actually a bug, but the default config includes stuff that depends on opencv, which doens't allow you to just do pip install thumbor and run it.

We need to upgrade tornado version in Pyvows.

Rewrite nose tests and add more tests to improve coverage.

Enable users to find focal points in the image by looking for glasses in OpenCV.

Create an OpenCV Image engine.

Having the encrypted one as default is more secure.

The urls should be:

/unsafe/300x300/smart/path/to/image.jpg

and

/1924931212k43j12fjsafiruqiu12041/path/to/image.jpg

Downloading even for files that have already been cached and stored in the file storage is very slow. On my local machine files that are only 55kb take over a second to load in the browser for each file.

URL generation and parsing needs work.