Huge feature (200+ commits from different developers). It completely changes layout of data (please expect very long DB migration process if you have a lot of tasks). The primary idea is to send data as zip chunks (e.g. 36 images in one chunk) or encoded video chunks and decode them on the client side. It helps to solve the problem with latency when you try to view a separate frame in the UI quickly (play mode).
Another important feature of the patch is to provide access to the original images. Thus for annotations the client uses compressed chunks but if you want to export a dataset Datumaro will use original chunks (but video will be decoded with original quality and encoded with maximum/optimal quality in any case).
* Don't use antd less (big memory consumtion during the build process)
* Fix AWS deployment guide
* fix a problem with proxy and long domain names
* remove sass loader for antd
* Removed less and less-loader.
* Simplified webpack config.
* Some debian package manager tweaks
By default, Ubuntu or Debian based "apt" or "apt-get" system installs recommended but not suggested packages .
By passing "--no-install-recommends" option, the user lets apt-get know not to consider recommended packages as a dependency to install.
This results in smaller downloads and installation of packages .
Refer to blog at [Ubuntu Blog](https://ubuntu.com/blog/we-reduced-our-docker-images-by-60-with-no-install-recommends) .
* Add image meta reading to voc
* Replace voc support in cvat
* Bump format version
* Materialize lazy transforms in voc export
* Store voc instance id as group id
* Add flat format import
* Add documentation
* Fix format name in doc
* Cuboid feature
* migration files
* Refactored cuboidShape
Fixed a bug where coloring by label would not update cuboids properly
Fixed a bug where the select points would not scale properly on initialization
* Removed math.js dependency
Implemented custom line intersection function
* new cvat formatting with labelled points
* Added MIT License to js files that were missing it
* Added simple constraints to the cuboids
* reverted commit for settings for vscode to hide local path
* fixed locking for cuboids
* fixed cuboid View when locked
* fixed occlusion view for cuboids
* Allow cuboid points to be outside the frame dimensions.
Signed-off-by: Tritin Truong <truongtritin98@gmail.com>
* Added stricter constraints on cuboid edges.
* Slightly stricter restrictions for edge case
* Cleaned up unused imports
* removed dashed lines on cuboids
* Moved projection lines to settings tab
* Fixed Cuboid shape buffer \
* Fix migrations (two 022 migrations after merge with the develop branch).
* Fix compatibility issues with auto segmentation.
* Grab points and update control scheme
* Greatly improved control scheme, fixed shape merging
Fixed Cuboid upload
* Fixed slight visual bug when dragging faces
* Some optimizations
* Hiding the grab point on creation
Small refactoring
* Fixed some cases where cuboid breaks
* Fixed upload for videos
* Removed perspective effects
* Made left back edge editable
* left back edge resizable
* fix statistics bug
* added toggles for the back edges
* Constraints for the back edges
* Fix creation bug
* Tightened creation constraints
* Fixing the code style
* updated message for invalid cuboids
* Code style
* More style fixes
* Codacy fixes
* added shift control for edges
* More Codacy fixes
* More Codacy fixes
* Double arrows for cursor
* Fix Drag bug
* More Codacy fixes
* Fix double quotes
* Fix camel case
* More camelcase fixes
* Generic object sink fixes
* Various codacy fixes
* Codacy
* Double quotes
* Fix migrations
* Updated shape creation
Fix jittering
* Adjusted constraints
* Codacy fixes
* Codacy fixes again
* Drawing cuboids from the top and bottom
* Codacy
* Resetting perspective on cuboids
* Choosing orientation of cuboids.
* Codacy fix
* Merge cleanup
* revert vs-code settings
* Update settings.json
Co-authored-by: timbowl <54648082+timbowl@users.noreply.github.com>
Co-authored-by: Nikita Manovich <40690625+nmanovic@users.noreply.github.com>
* Employ transforms and item wrapper
* Add image class and tests
* Add image info support to formats
* Fix cli
* Fix merge and voc converte
* Update remote images extractor
* Codacy
* Remove item name, require path in Image
* Merge images of dataset items
* Update tests
* Add image dir converter
* Update Datumaro format
* Update COCO format with image info
* Update CVAT format with image info
* Update TFrecord format with image info
* Update VOC formar with image info
* Update YOLO format with image info
* Update dataset manager bindings with image info
* Add image name to id transform
* Replace YOLO export and import in CVAT with Datumaro
* Employ transforms and item wrapper
* Add image class and tests
* Add image info support to formats
* Fix cli
* Fix merge and voc converte
* Update remote images extractor
* Codacy
* Remove item name, require path in Image
* Merge images of dataset items
* Update tests
* Add image dir converter
* Update Datumaro format
* Update COCO format with image info
* Update CVAT format with image info
* Update TFrecord format with image info
* Update VOC formar with image info
* Update YOLO format with image info
* Update dataset manager bindings with image info
* Add image name to id transform
* Fix coco export
Modification necessary for using CVAT from remote machines when accessing with FQDNs
See https://github.com/opencv/cvat/issues/1011#issue-542817055
and https://github.com/opencv/cvat/pull/1098
"I believe the reason for this is that sometimes if the port number is :80 and the URL is not in the LAN (:port), but instead it is a Fully Qualified Domain Name (:port), the port 80 is redundant (mydomain.com:80) and the errors arise."
* Replaced wget by curl
* Moved CI stuff into Dockerfile.ci
* Use docker-compose to run commnands inside docker (need environment variables)
* Added patool again (to support different archive formats)
* Roll back tensorflow version: 1.15 -> 1.13.1
Fixed https://github.com/opencv/cvat/issues/982
Fixed https://github.com/opencv/cvat/issues/1017
* datumaro install tensorflow 2.x now. It breaks automatic annotation
using TF.
* Follow redirects in curl (auto_segmentation)
* Added page_size parameter for all REST API methods which returns list of objects.
Also it is possible to specify page_size=all to return all elements.
* Updated changelog.md
* Add dataset export button for tasks in dashboard
* Fix downloading, shrink list of export formats
* Add strict export format check
* Add strict export format check
* Change REST api paths
* Move formats declarations to server,
* Supported git to create and sync
* Updated antd
* Updated icons
* Improved header
* Top bar for models & empty models list
* Removed one extra reducer and actions
* Removed one extra reducer and actions
* Crossplatform css
* Models reducers, some models actions, base for model list, imrovements
* Models list, ability to delete models
* Added ability to upload models
* Improved form, reinit models after create
* Removed some importants in css
* Model running dialog window, a lot of fixes
* Adding dump and load support for MOT CSV format.
* Updated test cases to use correct track annotations for MOT format.
* Removed behaviour of MOT loader which would duplicate the last track
shape prior to setting outside=True.
* Separated component user selector
* Change job assignee
* Basic create task window
* Bug fixes and refactoring
* Create task connected with a server
* Loading status for a button
* Reset loading on error response
* UI improvements
* Github/feedback/share window
* added handling of truncated and difficult attributes for pascal voc
loader/dumper
added descriptions of supported annotation formats
* added YOLO example
* made match_frame as Annotations method
changed 'image/source_id' field TF feature from int64 to string
(according to TF OD API dataset utlis)
* updated README
improved match_frame function
* added unit tests for dump/load
* Disable fix_segments_intersections for now
When the bounding boxes had intersections and were exported with the COCO JSON format they were often cut off. I commented out the line with the function fix_segments_intersections and replaced it with lines of that function. This helped with the bounding boxes and keeps the masks as they are created with CVAT. It is probably inconvenient for the user to get something fixed in the export without an active agreement of the user. Secondly letting a function automatically fix segments could result in a bad fix.
* Use fix_segments_intersections only with z-order
The fix_segments_intersections will only be used when the z-order flag is set. This is useful for bounding boxes or masks which don't need to be fixed. This fix was created according to Andrey Zhavoronkov's (@azhavoro) advice.
* Fix the problem with duplicated frames in case of "share".
* Fix a case when the code works incorrectly
/a/b/c
/a/b/c0
Previously only /a/b/c will be in output but should be both.
* Fix HTTP 400 error if together with vision data the user submit non-vision data (e.g. text files)
* Ignore SVG images because Pillow doesn't work with them.
* fill absent attributes by default values during annotation save
* fill absent attributes by default values during init from db
* fixed tests
* updated changelog, added some coments, minor fixes
* Added trivial login/logout/register
* Auth methods for REST API.
- api/v1/auth/login
- api/v1/auth/logout
For basic auth only:
- api/v1/auth/register
- api/v1/auth/password/*
* Add info about auth for REST API into CHANGELOG.md
* Add pylintrc for codacy, updated pylint and its dependecies.
* Add token authorization, renamed cvat.js to cvat-core in all places.
* Implemented register method in cvat-core
* Added first_name and last_name to RegisterSerializer.
* added tfrecord loader/dumper
* add comment
* remove unused import
* used the latest version of tensorflow(1.12.3) which supports cuda 9.0
updated cudnn library
install tensorflow by default and replace it by tensorflow-gpu in case
of cuda support is enabled
* Updated changelog
CVE-2019-12308 More information
moderate severity
Vulnerable versions: >= 2.1.0, < 2.1.9
Patched version: 2.1.9
An issue was discovered in Django 1.11 before 1.11.21, 2.1 before 2.1.9, and 2.2 before 2.2.2. The clickable Current URL value displayed by the AdminURLFieldWidget displays the provided value without validating it as a safe URL. Thus, an unvalidated value stored in the database, or a value provided as a URL query parameter payload, could result in an clickable JavaScript link.
WS-2019-0037 More information
moderate severity
Vulnerable versions: < 3.9.1
Patched version: 3.9.1
Django-Rest-Framework, before 3.9.1, has a XSS vulnerability caused by disabled autoescaping in the default DRF Browsable API view templates.
* Initial version of install guide for Ubuntu 18.04
* Description how to create install CVAT on Windows + fix for Windows 3rdparty patch.
* Write instructions for Mac OS.
* Clean up for README.md
* Update CHANGELOG.
* Fix typos which were found on code review.
* Fixed: single point wasn't created in right scenario, double press N (without any point) created a point
* Fixed: merge didn't create outside objects in some cases
* Fixed: split didn't work right
* Object selection propagated to next frame
Andrey Zhavoronkov
Nikita Manovich
Ben Hoff
zhiltsov-max
Boris Sekachev
Mathis Chenuet
Pratik Raj
Tritin Truong
Tritin Truong
dependabot[bot]
Jijoong Kim
LukeAI
Josh Bradley
Radhika
provider161
Christian
roho
TOsmanov
vugia truong
Maya
Maksim Markelov
Boris Sekachev
telenachos
Reza Malek
Ben Hoff
Johannes222
Naval Chand
Artyom Zankevich
JADG14
zliang7
Boris Sekachev
Eduardo
Satoshi Oikawa
Nikita Manovich
Santosh Thoduka
Happyzippy
kshramt
Julian Guarin
jrjbertram
Toni Kunic
aschernov
DanVev