All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- CVAT-3D: support lidar data on the server side (#2534)
- GPU support for Mask-RCNN and improvement in its deployment time (#2714)
- CVAT-3D: Load all frames corresponding to the job instance (#2645)
- Intelligent scissors with OpenCV javascript (#2689)
- CVAT-3D: Visualize 3D point cloud spaces in 3D View, Top View Side View and Front View (#2768)
- Inside Outside Guidence serverless function for interative segmentation
- Pre-built cvat_server and cvat_ui images were published on DockerHub (#2766)
- Project task subsets (#2774)
- Kubernetes templates and guide for their deployment (#1962)
- WiderFace format support (#2864)
- VGGFace2 format support (#2865)
- Backup/Restore guide (#2964)
- Label deletion from tasks and projects (#2881)
- CVAT-3D: Implemented initial cuboid placement in 3D View and select cuboid in Top, Side and Front views (#2891)
- Market-1501 format support (#2869)
- Ability of upload manifest for dataset with images (#2763)
- Annotations filters UI using react-awesome-query-builder (#1418)
- ICDAR format support (#2866)
- CLI - task list now returns a list of current tasks. (#2863)
- Updated HTTPS install README section (cleanup and described more robust deploy)
- Logstash is improved for using with configurable elasticsearch outputs (#2531)
- Bumped nuclio version to 1.5.16 (#2578)
- All methods for interative segmentation accept negative points as well
- Persistent queue added to logstash (#2744)
- Improved maintanance of popups visibility (#2809)
- Image visualizations settings on canvas for faster access (#2872)
- Better scale management of left panel when screen is too small (#2880)
- Improved error messages for annotation import (#2935)
- Using manifest support instead video meta information and dummy chunks (#2763)
- More robust execution of nuclio GPU functions by limiting the GPU memory consumption per worker (#2714)
- Kibana startup initialization (#2659)
- The cursor jumps to the end of the line when renaming a task (#2669)
- SSLCertVerificationError when remote source is used (#2683)
- Fixed filters select overflow (#2614)
- Fixed tasks in project autoannotation (#2725)
- Cuboids are missed in annotations statistics (#2704)
- The list of files attached to the task is not displayed (<https://github.com/openvinotoolkit/cvat/pul
- A couple of css-related issues (top bar disappear, wrong arrow position on collapse elements) (#2736)
- Issue with point region doesn't work in Firefox (#2727)
- Fixed cuboid perpective change (#2733)
- Annotation page popups (ai tools, drawing) reset state after detecting, tracking, drawing (#2780)
- Polygon editing using trailing point (#2808)
- Updated the path to python for DL models inside automatic annotation documentation (#2847)
- Fixed of receiving function variable (#2860)
- Shortcuts with CAPSLOCK enabled and with non-US languages activated (#2872)
- Prevented creating several issues for the same object (#2868)
- Fixed label editor name field validator (#2879)
- An error about track shapes outside of the task frames during export (#2890)
- Fixed project search field updating (#2901)
- Fixed export error when invalid polygons are present in overlapping frames (#2852)
- Fixed image quality option for tasks created from images (#2963)
- Incorrect text on the warning when specifying an incorrect link to the issue tracker (#2971)
- Updating label attributes when label contains number attributes (#2969)
- Memory consumption for the task creation process (#2582)
- Frame preloading (#2608)
- Project cannot be removed from the project page (#2626)
- GPU support and improved documentation for auto annotation (#2546)
- Manual review pipeline: issues/comments/workspace (#2357)
- Basic projects implementation (#2255)
- Documentation on how to mount cloud starage(AWS S3 bucket, Azure container, Google Drive) as FUSE (#2377)
- Ability to work with share files without copying inside (#2377)
- Tooltips in label selectors (#2509)
- Page redirect after login using
next
query parameter (#2527) - ImageNet format support (#2376)
- CamVid format support (#2559)
- PATCH requests from cvat-core submit only changed fields (#2445)
- deploy.sh in serverless folder is seperated into deploy_cpu.sh and deploy_gpu.sh (#2546)
- Bumped nuclio version to 1.5.8
- Migrated to Antd 4.9 (#2536)
- Fixed FastRCNN inference bug for images with 4 channels i.e. png (#2546)
- Django templates for email and user guide (#2412)
- Saving relative paths in dummy chunks instead of absolute (#2424)
- Objects with a specific label cannot be displayed if at least one tag with the label exist (#2435)
- Wrong attribute can be removed in labels editor (#2436)
- UI fails with the error "Cannot read property 'label' of undefined" (#2442)
- Exception: "Value must be a user instance" (#2441)
- Reset zoom option doesn't work in tag annotation mode (#2443)
- Canvas is busy error (#2437)
- Projects view layout fix (#2503)
- Fixed the tasks view (infinite loading) when it is impossible to get a preview of the task (#2504)
- Empty frames navigation (#2505)
- TypeError: Cannot read property 'toString' of undefined (#2517)
- Extra shapes are drawn after Esc, or G pressed while drawing a region in grouping (#2507)
- Reset state (reviews, issues) after logout or changing a job (#2525)
- TypeError: Cannot read property 'id' of undefined when updating a task (#2544)
- Ability to login into CVAT-UI with token from api/v1/auth/login (#2234)
- Added layout grids toggling ('ctrl + alt + Enter')
- Added password reset functionality (#2058)
- Ability to work with data on the fly (#2007)
- Annotation in process outline color wheel (#2084)
- On the fly annotation using DL detectors (#2102)
- Displaying automatic annotation progress on a task view (#2148)
- Automatic tracking of bounding boxes using serverless functions (#2136)
- [Datumaro] CLI command for dataset equality comparison (#1989)
- [Datumaro] Merging of datasets with different labels (#2098)
- Add FBRS interactive segmentation serverless function (#2094)
- Ability to change default behaviour of previous/next buttons of a player. It supports regular navigation, searching a frame according to annotations filters and searching the nearest frame without any annotations (#2221)
- MacOS users notes in CONTRIBUTING.md
- Ability to prepare meta information manually (#2217)
- Ability to upload prepared meta information along with a video when creating a task (#2217)
- Optional chaining plugin for cvat-canvas and cvat-ui (#2249)
- MOTS png mask format support (#2198)
- Ability to correct upload video with a rotation record in the metadata (#2218)
- User search field for assignee fields (#2370)
- Support of mxf videos (#2514)
- UI models (like DEXTR) were redesigned to be more interactive (#2054)
- Used Ubuntu:20.04 as a base image for CVAT Dockerfile (#2101)
- Right colors of label tags in label mapping when a user runs automatic detection (#2162)
- Nuclio became an optional component of CVAT (#2192)
- A key to remove a point from a polyshape [Ctrl => Alt] (#2204)
- Updated
docker-compose
file version from2.3
to3.3
(#2235) - Added auto inference of url schema from host in CLI, if provided (#2240)
- Track frames in skips between annotation is presented in MOT and MOTS formats are marked
outside
(#2198) - UI packages installation with
npm ci
instead ofnpm install
(#2350)
- Removed Z-Order flag from task creation process
- Fixed multiple errors which arises when polygon is of length 5 or less (#2100)
- Fixed task creation from PDF (#2141)
- Fixed CVAT format import for frame stepped tasks (#2151)
- Fixed the reading problem with large PDFs (#2154)
- Fixed unnecessary pyhash dependency (#2170)
- Fixed Data is not getting cleared, even after deleting the Task from Django Admin App(#1925)
- Fixed blinking message: "Some tasks have not been showed because they do not have any data" (#2200)
- Fixed case when a task with 0 jobs is shown as "Completed" in UI (#2200)
- Fixed use case when UI throws exception: Cannot read property 'objectType' of undefined #2053 (#2203)
- Fixed use case when logs could be saved twice or more times #2202 (#2203)
- Fixed issues from #2112 (#2217)
- Git application name (renamed to dataset_repo) (#2243)
- A problem in exporting of tracks, where tracks could be truncated (#2129)
- Fixed CVAT startup process if the user has
umask 077
in .bashrc file (#2293) - Exception: Cannot read property "each" of undefined after drawing a single point (#2307)
- Cannot read property 'label' of undefined (Fixed?) (#2311)
- Excluded track frames marked
outside
inCVAT for Images
export (#2345) - 'List of tasks' Kibana visualization (#2361)
- An error on exporting not
jpg
orpng
images in TF Detection API format (openvinotoolkit/datumaro#35)
- Siammask tracker as DL serverless function (#1988)
- [Datumaro] Added model info and source info commands (#1973)
- [Datumaro] Dataset statistics (#1668)
- Ability to change label color in tasks and predefined labels (#2014)
- [Datumaro] Multi-dataset merge (#1695)
- Ability to configure email verification for new users (#1929)
- Link to django admin page from UI (#2068)
- Notification message when users use wrong browser (#2070)
- Shape coordinates are rounded to 2 digits in dumped annotations (#1970)
- COCO format does not produce polygon points for bbox annotations (#1953)
- Issue loading openvino models for semi-automatic and automatic annotation (#1996)
- Basic functions of CVAT works without activated nuclio dashboard
- Fixed a case in which exported masks could have wrong color order (#2032)
- Fixed error with creating task with labels with the same name (#2031)
- Django RQ dashboard view (#2069)
- Object's details menu settings (#2084)
- DL models as serverless functions (#1767)
- Source type support for tags, shapes and tracks (#1192)
- Source type support for CVAT Dumper/Loader (#1192)
- Intelligent polygon editing (#1921)
- Support creating multiple jobs for each task through python cli (#1950)
- python cli over https (#1942)
- Error message when plugins weren't able to initialize instead of infinite loading (#1966)
- Ability to change user password (#1954)
- Smaller object details (#1877)
COCO
format does not convert bboxes to polygons on export (#1953)- It is impossible to submit a DL model in OpenVINO format using UI. Now you can deploy new models on the server using serverless functions (#1767)
- Files and folders under share path are now alphabetically sorted
- Removed OpenVINO and CUDA components because they are not necessary anymore (#1767)
- Removed the old UI code (#1964)
- Some objects aren't shown on canvas sometimes. For example after propagation on of objects is invisible (#1834)
- CVAT doesn't offer to restore state after an error (#1874)
- Cannot read property 'shapeType' of undefined because of zOrder related issues (#1874)
- Cannot read property 'pinned' of undefined because of zOrder related issues (#1874)
- Do not iterate over hidden objects in aam (which are invisible because of zOrder) (#1874)
- Cursor position is reset after changing a text field (#1874)
- Hidden points and cuboids can be selected to be groupped (#1874)
outside
annotations should not be in exported images (#1620)CVAT for video format
import error with interpolation (#1893)Image compression
definition mismatch (#1900)- Points are dublicated during polygon interpolation sometimes (#1892)
- When redraw a shape with activated autobordering, previous points are visible (#1892)
- No mapping between side object element and context menu in some attributes (#1923)
- Interpolated shapes exported as
keyframe = True
(#1937) - Stylelint filetype scans (#1952)
- Fixed toolip closing issue (#1955)
- Clearing frame cache when close a task (#1966)
- Increase rate of throttling policy for unauthenticated users (#1969)
- Throttling policy for unauthenticated users (#1531)
- Added default label color table for mask export (#1549)
- Added environment variables for Redis and Postgres hosts for Kubernetes deployment support (#1641)
- Added visual identification for unavailable formats (#1567)
- Shortcut to change color of an activated shape in new UI (Enter) (#1683)
- Shortcut to switch split mode (#1683)
- Built-in search for labels when create an object or change a label (#1683)
- Better validation of labels and attributes in raw viewer (#1727)
- ClamAV antivirus integration (#1712)
- Added canvas background color selector (#1705)
- SCSS files linting with Stylelint tool (#1766)
- Supported import and export or single boxes in MOT format (#1764)
- [Datumaro] Added
stats
command, which shows some dataset statistics like image mean and std (#1734) - Add option to upload annotations upon task creation on CLI
- Polygon and polylines interpolation (#1571)
- Ability to redraw shape from scratch (Shift + N) for an activated shape (#1571)
- Highlights for the first point of a polygon/polyline and direction (#1571)
- Ability to change orientation for poylgons/polylines in context menu (#1571)
- Ability to set the first point for polygons in points context menu (#1571)
- Added new tag annotation workspace (#1570)
- Appearance block in attribute annotation mode (#1820)
- Keyframe navigations and some switchers in attribute annotation mode (#1820)
- [Datumaro] Added
convert
command to convert datasets directly (#1837) - [Datumaro] Added an option to specify image extension when exporting datasets (#1799)
- [Datumaro] Added image copying when exporting datasets, if possible (#1799)
- Removed information about e-mail from the basic user information (#1627)
- Update https install manual. Makes it easier and more robust. Includes automatic renewing of lets encrypt certificates.
- Settings page move to the modal. (#1705)
- Implemented import and export of annotations with relative image paths (#1463)
- Using only single click to start editing or remove a point (#1571)
- Added support for attributes in VOC XML format (#1792)
- Added annotation attributes in COCO format (#1782)
- Colorized object items in the side panel (#1753)
- [Datumaro] Annotation-less files are not generated anymore in COCO format, unless tasks explicitly requested (#1799)
- Problem with exported frame stepped image task (#1613)
- Fixed dataset filter item representation for imageless dataset items (#1593)
- Fixed interpreter crash when trying to import
tensorflow
with no AVX instructions available (#1567) - Kibana wrong working time calculation with new annotation UI use (#1654)
- Wrong rexex for account name validation (#1667)
- Wrong description on register view for the username field (#1667)
- Wrong resolution for resizing a shape (#1667)
- React warning because of not unique keys in labels viewer (#1727)
- Fixed issue tracker (#1705)
- Fixed canvas fit after sidebar open/close event (#1705)
- A couple of exceptions in AAM related with early object activation (#1755)
- Propagation from the latest frame (#1800)
- Number attribute value validation (didn't work well with floats) (#1800)
- Logout doesn't work (#1812)
- Annotations aren't updated after reopening a task (#1753)
- Labels aren't updated after reopening a task (#1753)
- Canvas isn't fitted after collapsing side panel in attribute annotation mode (#1753)
- Error when interpolating polygons (#1878)
- SQL injection in Django
CVE-2020-9402
(#1657)
- cvat-ui: cookie policy drawer for login page (#1511)
datumaro_project
export format (#1352)- Ability to configure user agreements for the user registration form (#1464)
- Cuboid interpolation and cuboid drawing from rectangles (#1560)
- Ability to configure custom pageViewHit, which can be useful for web analytics integration (#1566)
- Ability to configure access to the analytics page based on roles (#1592)
- Downloaded file name in annotations export became more informative (#1352)
- Added auto trimming for trailing whitespaces style enforcement (#1352)
- REST API: updated
GET /task/<id>/annotations
: parameters areformat
,filename
(now optional),action
(optional) (#1352) - REST API: removed
dataset/formats
, changed format ofannotation/formats
(#1352) - Exported annotations are stored for N hours instead of indefinitely (#1352)
- Formats: CVAT format now accepts ZIP and XML (#1352)
- Formats: COCO format now accepts ZIP and JSON (#1352)
- Formats: most of formats renamed, no extension in title (#1352)
- Formats: definitions are changed, are not stored in DB anymore (#1352)
- cvat-core: session.annotations.put() now returns ids of added objects (#1493)
- Images without annotations now also included in dataset/annotations export (#525)
annotation
application is replaced withdataset_manager
(#1352)_DATUMARO_INIT_LOGLEVEL
env. variable is removed in favor of regular--loglevel
cli parameter (#1583)
- Categories for empty projects with no sources are taken from own dataset (#1352)
- Added directory removal on error during
extract
command (#1352) - Added debug error message on incorrect XPath (#1352)
- Exporting frame stepped task (<#1294, https://github.com/opencv/cvat/issues/1334>)
- Fixed broken command line interface for
cvat
export format in Datumaro (#1494) - Updated Rest API document, Swagger document serving instruction issue (#1495)
- Fixed cuboid occluded view (#1500)
- Non-informative lock icon (#1434)
- Sidebar in AAM has no hide/show button (#1420)
- Task/Job buttons has no "Open in new tab" option (#1419)
- Delete point context menu option has no shortcut hint (#1416)
- Fixed issue with unnecessary tag activation in cvat-canvas (#1540)
- Fixed an issue with large number of instances in instance mask (#1539)
- Fixed full COCO dataset import error with conflicting labels in keypoints and detection (#1548)
- Fixed COCO keypoints skeleton parsing and saving (#1539)
tf.placeholder() is not compatible with eager execution
exception for auto_segmentation (#1562)- Canvas cannot be moved with move functionality on left mouse key (#1573)
- Deep extreme cut request is sent when draw any shape with Make AI polygon option enabled (#1573)
- Fixed an error when exporting a task with cuboids to any format except CVAT (#1577)
- Synchronization with remote git repo (#1582)
- A problem with mask to polygons conversion when polygons are too small (#1581)
- Unable to upload video with uneven size (#1594)
- Fixed an issue with
z_order
having no effect on segmentations (#1589)
- Permission group whitelist check for analytics view (#1608)
- Re-Identification algorithm to merging bounding boxes automatically to the new UI (#1406)
- Methods
import
andexport
to import/export raw annotations for Job and Task incvat-core
(#1406) - Versioning of client packages (
cvat-core
,cvat-canvas
,cvat-ui
). Initial versions are set to 1.0.0 (#1448) - Cuboids feature was migrated from old UI to new one. (#1451)
- Annotation convertation utils, currently supported natively via Datumaro framework (#1477)
- Auto annotation, TF annotation and Auto segmentation apps (#1409)
- Import works with truncated images now: "OSError:broken data stream" on corrupt images (#1430)
- Hide functionality (H) doesn't work (#1445)
- The highlighted attribute doesn't correspond to the chosen attribute in AAM (#1445)
- Inconvinient image shaking while drawing a polygon (hold Alt key during drawing/editing/grouping to drag an image) (#1445)
- Filter property "shape" doesn't work and extra operator in description (#1445)
- Block of text information doesn't disappear after deactivating for locked shapes (#1445)
- Annotation uploading fails in annotation view (#1445)
- UI freezes after canceling pasting with escape (#1445)
- Duplicating keypoints in COCO export (#1435)
- CVAT new UI: add arrows on a mouse cursor (#1391)
- Delete point bug (in new UI) (#1440)
- Fix apache startup after PC restart (#1467)
- Open task button doesn't work (#1474)
- Special behaviour for attribute value
__undefined__
(invisibility, no shortcuts to be set in AAM) - Dialog window with some helpful information about using filters
- Ability to display a bitmap in the new UI
- Button to reset colors settings (brightness, saturation, contrast) in the new UI
- Option to display shape text always
- Dedicated message with clarifications when share is unmounted (#1373)
- Ability to create one tracked point (#1383)
- Ability to draw/edit polygons and polylines with automatic bordering feature (#1394)
- Tutorial: instructions for CVAT over HTTPS
- Deep extreme cut (semi-automatic segmentation) to the new UI (#1398)
- Increase preview size of a task till 256, 256 on the server
- Public ssh-keys are displayed in a dedicated window instead of console when create a task with a repository
- React UI is the primary UI
- Cleaned up memory in Auto Annotation to enable long running tasks on videos
- New shape is added when press
esc
when drawing instead of cancellation - Dextr segmentation doesn't work.
FileNotFoundError
during dump after moving format files- CVAT doesn't append outside shapes when merge polyshapes in old UI
- Layout sometimes shows double scroll bars on create task, dashboard and settings pages
- UI fails after trying to change frame during resizing, dragging, editing
- Hidden points (or outsided) are visible after changing a frame
- Merge is allowed for points, but clicks on points conflict with frame dragging logic
- Removed objects are visible for search
- Add missed task_id and job_id fields into exception logs for the new UI (#1372)
- UI fails when annotations saving occurs during drag/resize/edit (#1383)
- Multiple savings when hold Ctrl+S (a lot of the same copies of events were sent with the same working time) (#1383)
- UI doesn't have any reaction when git repos synchronization failed (#1383)
- Bug when annotations cannot be saved after (delete - save - undo - save) (#1383)
- VOC format exports Upper case labels correctly in lower case (#1379)
- Fixed polygon exporting bug in COCO dataset (#1387)
- Task creation from remote files (#1392)
- Job cannot be opened in some cases when the previous job was failed during opening (#1403)
- Deactivated shape is still highlighted on the canvas (#1403)
- AttributeError: 'tuple' object has no attribute 'read' in ReID algorithm (#1403)
- Wrong semi-automatic segmentation near edges of an image (#1403)
- Git repos paths (#1400)
- Uploading annotations for tasks with multiple jobs (#1396)
- Data streaming using chunks (#1007)
- New UI: showing file names in UI (#1311)
- New UI: delete a point from context menu (#1292)
- Git app cannot clone a repository (#1330)
- New UI: preview position in task details (#1312)
- AWS deployment (#1316)
- VOC task export now does not use official label map by default, but takes one from the source task to avoid primary-class and class part name clashing (#1275)
- File names in LabelMe format export are no longer truncated (#1259)
occluded
andz_order
annotation attributes are now correctly passed to Datumaro (#1271)- Annotation-less tasks now can be exported as empty datasets in COCO (#1277)
- Frame name matching for video annotations import -
allowed
frame_XXXXXX[.ext]
format (#1274)
- Bump acorn from 6.3.0 to 6.4.1 in /cvat-ui (#1270)
- Server only support for projects. Extend REST API v1 (/api/v1/projects*)
- Ability to get basic information about users without admin permissions (#750)
- Changed REST API: removed PUT and added DELETE methods for /api/v1/users/ID
- Mask-RCNN Auto Annotation Script in OpenVINO format
- Yolo Auto Annotation Script
- Auto segmentation using Mask_RCNN component (Keras+Tensorflow Mask R-CNN Segmentation)
- REST API to export an annotation task (images + annotations)
- Datumaro - a framework to build, analyze, debug and visualize datasets
- Text Detection Auto Annotation Script in OpenVINO format for version 4
- Added in OpenVINO Semantic Segmentation for roads
- Ability to visualize labels when using Auto Annotation runner
- MOT CSV format support (#830)
- LabelMe format support (#844)
- Segmentation MASK format import (as polygons) (#1163)
- Git repositories can be specified with IPv4 address (#827)
- page_size parameter for all REST API methods
- React & Redux & Antd based dashboard
- Yolov3 interpretation script fix and changes to mapping.json
- YOLO format support (#1151)
- Added support for OpenVINO 2020
- Exception in Git plugin #826
- Label ids in TFrecord format now start from 1 #866
- Mask problem in COCO JSON style #718
- Datasets (or tasks) can be joined and split to subsets with Datumaro #791
- Output labels for VOC format can be specified with Datumaro #942
- Annotations can be filtered before dumping with Datumaro #994
- Frozen version of scikit-image==0.15 in requirements.txt because next releases don't support Python 3.5
- Integration with Zenodo.org (DOI)
- A converter to YOLO format
- Installation guide
- Linear interpolation for a single point
- Video frame filter
- Running functional tests for REST API during a build
- Admins are no longer limited to a subset of python commands in the auto annotation application
- Remote data source (list of URLs to create an annotation task)
- Auto annotation using Faster R-CNN with Inception v2 (utils/open_model_zoo)
- Auto annotation using Pixel Link mobilenet v2 - text detection (utils/open_model_zoo)
- Ability to create a custom extractors for unsupported media types
- Added in PDF extractor
- Added in a command line model manager tester
- Ability to dump/load annotations in several formats from UI (CVAT, Pascal VOC, YOLO, MS COCO, png mask, TFRecord)
- Auth for REST API (api/v1/auth/): login, logout, register, ...
- Preview for the new CVAT UI (dashboard only) is available: http://localhost:9080/
- Added command line tool for performing common task operations (/utils/cli/)
- Outside and keyframe buttons in the side panel for all interpolation shapes (they were only for boxes before)
- Improved error messages on the client side (#511)
- "Flip images" has been removed. UI now contains rotation features.
- Incorrect width of shapes borders in some cases
- Annotation parser for tracks with a start frame less than the first segment frame
- Interpolation on the server near outside frames
- Dump for case when task name has a slash
- Auto annotation fail for multijob tasks
- Installation of CVAT with OpenVINO on the Windows platform
- Background color was always black in utils/mask/converter.py
- Exception in attribute annotation mode when a label are switched to a value without any attributes
- Handling of wrong labelamp json file in auto annotation (#554)
- No default attributes in dumped annotation (#601)
- Required field "Frame Filter" on admin page during a task modifying (#666)
- Dump annotation errors for a task with several segments (#610, #500)
- Invalid label parsing during a task creating (#628)
- Button "Open Task" in the annotation view
- Creating a video task with 0 overlap
- Upgraded Django, djangorestframework, and other packages
- Fixed interaction with the server share in the auto annotation plugin
- JavaScript syntax incompatibility with Google Chrome versions less than 72
- OpenVINO auto annotation: it is possible to upload a custom model and annotate images automatically.
- Ability to rotate images/video in the client part (Ctrl+R, Shift+Ctrl+R shortcuts) (#305)
- The ReID application for automatic bounding box merging has been added (#299)
- Keyboard shortcuts to switch next/previous default shape type (box, polygon etc) [Alt + <, Alt + >] (#316)
- Converter for VOC now supports interpolation tracks
- REST API (/api/v1/*, /api/docs)
- Semi-automatic semantic segmentation with the Deep Extreme Cut work
- Propagation setup has been moved from settings to bottom player panel
- Additional events like "Debug Info" or "Fit Image" have been added for analitics
- Optional using LFS for git annotation storages (#314)
- "Flip images" flag in the create task dialog will be removed. Rotation functionality in client part have been added instead.
- Django 2.1.5 (security fix, https://nvd.nist.gov/vuln/detail/CVE-2019-3498)
- Several scenarious which cause code 400 after undo/redo/save have been fixed (#315)
- Ability to copy Object URL and Frame URL via object context menu and player context menu respectively.
- Ability to change opacity for selected shape with help "Selected Fill Opacity" slider.
- Ability to remove polyshapes points by double click.
- Ability to draw/change polyshapes (except for points) by slip method. Just press ENTER and moving a cursor.
- Ability to switch lock/hide properties via label UI element (in right menu) for all objects with same label.
- Shortcuts for outside/keyframe properties
- Support of Intel OpenVINO for accelerated model inference
- Tensorflow annotation now works without CUDA. It can use CPU only. OpenVINO and CUDA are supported optionally.
- Incremental saving of annotations.
- Tutorial for using polygons (screencast)
- Silk profiler to improve development process
- Admin panel can be used to edit labels and attributes for annotation tasks
- Analytics component to manage a data annotation team, monitor exceptions, collect client and server logs
- Changeable job and task statuses (annotation, validation, completed). A job status can be changed manually, a task status is computed automatically based on job statuses (#153)
- Backlink to a task from its job annotation view (#156)
- Buttons lock/hide for labels. They work for all objects with the same label on a current frame (#116)
- Polyshape editing method has been improved. You can redraw part of shape instead of points cloning.
- Unified shortcut (Esc) for close any mode instead of different shortcuts (Alt+N, Alt+G, Alt+M etc.).
- Dump file contains information about data source (e.g. video name, archive name, ...)
- Update requests library due to https://nvd.nist.gov/vuln/detail/CVE-2018-18074
- Per task/job permissions to create/access/change/delete tasks and annotations
- Documentation was improved
- Timeout for creating tasks was increased (from 1h to 4h) (#136)
- Drawing has become more convenience. Now it is possible to draw outside an image. Shapes will be automatically truncated after drawing process (#202)
- Performance bottleneck has been fixed during you create new objects (draw, copy, merge etc).
- Label UI elements aren't updated after changelabel.
- Attribute annotation mode can use invalid shape position after resize or move shapes.
- Labels order is preserved now (#242)
- Uploading large XML files (#123)
- Django vulnerability (#121)
- Grammatical cleanup of README.md (#107)
- Dashboard loading has been accelerated (#156)
- Text drawing outside of a frame in some cases (#202)
- New annotation shapes: polygons, polylines, points
- Undo/redo feature
- Grid to estimate size of objects
- Context menu for shapes
- A converter to PASCAL VOC format
- A converter to MS COCO format
- A converter to mask format
- License header for most of all files
- .gitattribute to avoid problems with bash scripts inside a container
- CHANGELOG.md itself
- Drawing size of a bounding box during resize
- Color by instance, group, label
- Group objects
- Object propagation on next frames
- Full screen view
- Documentation, screencasts, the primary screenshot
- Content-type for save_job request is application/json
- Player navigation if the browser's window is scrolled
- Filter doesn't support dash (-)
- Several memory leaks
- Inconsistent extensions between filenames in an annotation file and real filenames
- 7z archive support when creating a task
- .vscode/launch.json file for developing with VS code
- #14: docker-compose down command as written in the readme does not remove volumes
- #15: all checkboxes in temporary attributes are checked when reopening job after saving the job
- #18: extend CONTRIBUTING.md
- #19: using the same attribute for label twice -> stuck
- More strict verification for labels with attributes
- Links on a screenshot, documentation, screencasts into README.md
- CONTRIBUTORS.md
- GitHub documentation
- Initial version
## [Unreleased]
### Added
-
### Changed
-
### Deprecated
-
### Removed
-
### Fixed
-
### Security
-