Changelog for package jsk_arc2017_common

4.3.0 (2021-07-14)

• Merge pull request #2704 from pazeshun/install_pick_re-experiment
• [jsk_arc2017_common] Add script to install pick_re-experiment (experiment imitating ARC2017 pick competition)
• Merge pull request #2709 from pazeshun/add-baxtergv6_apc2016
• Enable to change proba_threshold in object_segmentation_3d.launch
• Stop downloading unused model checkpoints
• Contributors: Kentaro Wada, Shingo Kitagawa, Shun Hasegawa

4.2.1 (2019-04-18)

• ignore fcn32s.npz
• Add WeightChangeDetector to tell weight change without candidates
• Fix path of dataset and download raw data automatically
• Update for updated Object.msg
• Flip model's axis to make them face to z
• Object 3d annotation to multi-view log data
• Update candidates_publisher.py
• Update candidates_publisher.py
• Fix unreasonable ~json_dir param to candidates_publisher
• Skip empty bin while visualizing JSON
• Don't ignore fcn32s.npz to remove it by git clean
• Modify weight of poland_spring_water
• use timer in candidates_publisher
• Contributors: Kentaro Wada, Shingo Kitagawa, Shun Hasegawa, Yuto Uchimi

4.2.0 (2017-11-01)

• Add README
• Add .gitignore
• Install item_data via a script for few-shot segmentation
• Contributors: Kentaro Wada

4.1.9 (2017-10-28)

4.1.8 (2017-10-26)

4.1.7 (2017-10-26)

• 4.1.6
• Update CHANGELOG.rst
• 4.1.5
• Update CHANGELOG.rst
• Contributors: Kentaro Wada

4.1.4 (2017-10-15)

• Skip ResourceNotFound error to pass build on build.ros.org
• Contributors: Kentaro Wada

4.1.6 (2017-10-24)

4.1.5 (2017-10-23)

• 4.1.4
• Update CHANGELOG
• Skip ResourceNotFound error to pass build on build.ros.org
• Contributors: Kentaro Wada

4.1.3 (2017-10-12)

4.1.2 (2017-10-11)

• Create fcn32s.npz symlink by install_data.py In order to avoid error in bloom.
• Contributors: Kentaro Wada

4.1.1 (2017-10-10)

• Equalize weight_candidates_refiner in jsk_apc to upstream PRs
• Create ItemData by taking photos
• Contributors: Kentaro Wada, Shun Hasegawa

4.1.0 (2017-08-12)

• Warn about nan value
• Publish nan value when unstable by EkEwIDriver
• use 1B2 instead of K3 for test
• use sample_pick_task json for sample launch
• fix bug in cardboard ids 1A5 is bigger than 1B2, but 1A5 is never used for cardboard_C and 1B2 for cardboard_B
• Contributors: Kentaro Wada, Shingo Kitagawa

4.0.9 (2017-07-29)

• Fix typo in visualize_object.py
• Contributors: Kentaro Wada

4.0.8 (2017-07-29)

4.0.7 (2017-07-28)

• fix visualize_object
• Contributors: Kentaro Wada

4.0.6 (2017-07-28)

4.0.5 (2017-07-28)

4.0.4 (2017-07-27)

4.0.3 (2017-07-27)

• Set default model
• change suction graspability of avery_binder
• fix nosetest path in jsk_arc2017_common
• Show with red color for unknown objects in list_label_names
• Add get_known_object_names and its tests
• Fixes for unknown objects
• Use fcn32s.npz in models dir
• Support unknown objects disclosed just before the challenge This requires https://github.com/start-jsk/jsk_apc/issues/2414#issuecomment-317935277
• Contributors: Kentaro Wada, Naoya Yamaguchi, Shingo Kitagawa

4.0.2 (2017-07-27)

• change pinch graspability of white_facecloth
• publish bin_contents in json_saver
• add Content and ContentArray msg
• change reynolds wrap graspability
• Contributors: Naoya Yamaguchi, Shingo Kitagawa

4.0.1 (2017-07-26)

• Reset pub_imgs
• Classify object in grasped region
• fix apply_mask_to_proba_image.py:32:5: E303 too many blank lines (2) ?
• create apply_mask_to_proba_image.py
• create image_buffer.py
• 148000 iterations with Datasetv3 ` % ./evaluate.py logs/fcn32s_CFG-009_DATASET-v3_OPTIMIZER-Adam_WEIGHT_DECAY-0.0005_LR-1e-05_MAX_ITERATION-150000_VCS-c0bff11_TIME-20170723-112528/models/fcn32s_dataset_v3_cfg009_20170724.npz Accuracy: 93.7259 AccClass: 92.3574 Mean IoU: 84.6101 Fwav Acc: 88.3183`
• pinch graspability 2 -> 3
• Contributors: Kei Okada, Kentaro Wada, Naoya Yamaguchi, Yuto Uchimi

4.0.0 (2017-07-24)

• Fix workorder in shared shelf-bin (B)
• Fix suction for pie plates
• Enhance object location display
• fix graspability of some items
• Fix graspability of speed stick
• FCN32s trained using natural dataset (datasetV3)
• add grasp_style_server.py
• add GetGraspStyle service
• sort work order by graspability
• add list_graspability script
• add func get_object_graspability()
• Fixed training of FCN32s using stacking data augmentation
• No use of ConnectionBasedTransport in WeightCanditatesRefiner
• Fix old timestamp in EkEwIDriver output
• Always subscribe weight scale in weight_candidates_refiner
• Use class segmentation in known objects Because we changed the strategy to handle the unknown (newly passed) objects.
• Mark ordered objects with red circle in VisualizeJSON
• Fix for pep8
• Update ekew_i_driver.py
• Update ekew_i_driver.py
• Update ekew_i_driver.py
• change topic name
• always publish raw weight value
• Fix typo in weight_candidates_refiner.py
• publish prev_weight_sum for debugging (#2322)
  • publish prev_weight_sum for debugging
  • Update weight_candidates_refiner.py
  • Fix typo
  • Update weight_candidates_refiner.py
• sort work order by object weight, pick lighter one
• add get_object_weights() in jsk_arc2017_common
• move object weight yaml to config dir
• Update README.md
• Disable of downloading old chainer models
• Create dataset V2
• Rename scripts annotate_dataset2d.py, view_dataset2d.py
• Contributors: Kentaro Wada, Naoya Yamaguchi, Shingo Kitagawa, Shun Hasegawa

3.3.0 (2017-07-15)

• Add script to visualize annotated 2d dataset
• fix E271 multiple spaces after keyword ERROR.…
• Detect serial blocked and restart
• Update to support multi shelf bins
• Add README to annotate_2d_dataset
• Rename to annotate_2d_dataset.py
• Publish scenes and view frame of DatasetV3 in ROS
• merge json_generator into one program (#2270)
• Fix for flake8
• Memoize result of visualize_json
• refine weight_candidates_refiner node
• publish -1 when scale is disabled
• remove unused launch
• rename to weight_candidates_refiner node
• add use_topic and input_candidates args
• sub candidates in scale object estimation node
• publish WeightStamped from scale node
• add Weight and WeightStamped msg
• replace bg_label by ignore_labels
• use arc2017 object_segmentation_3d in stow task
• ad ignore_labels in label_to_cpi
• add USE_PCA argment in object_segmentation_3d.launch
• Contributors: Kei Okada, Kentaro Wada, Naoya Yamaguchi, Shingo Kitagawa, Shun Hasegawa

3.2.0 (2017-07-06)

• add object_classification with FCN launch
• add doc, sample and test for candidates_publisher
• update Label msg API follow https://github.com/jsk-ros-pkg/jsk_recognition/pull/2143/commits/109c73fac35f1cdaa13fd31273ca166b2bcbfce9
• add candidates_publisher node
• Create object_segmentation_3d.launch in jsk_arc2017_common
• Semantic segmentation of unknown objects
• Use simlink to scales
• Add udev rule for scale
• Support json with no boxes in visualize_json
• Update doc for visualize_json.py
• Subscribe json_dir input topic in visualize_json
• Publish json_dir in json_saver.py
• Visualize json (item_location/order) for ARC tasks
• copy location and order json in save dir at first
• json_saver supports pick task
• Contributors: Kentaro Wada, Shingo Kitagawa, Shun Hasegawa

3.1.0 (2017-06-30)

• Add mesh models for 36 objects
• update UpdateJSON and replace SaveJSON by Trigger
• add json_saver.py
• Refactor yaml file format
• Fix format
• Add thread lock to estimate_object_by_scale
• Change init of object estimation to srv
• Add object estimation by scale
• and_scale_rosserial -> ekew_i_driver
• Add object weight data
• remove non-item label in json generator
• update sample_pick_task json
• update pick_json_generator for new pick env
• update work_order_publisher for new shelf
• Place location config files in jsk_arc2017_baxter
• add publish_tote_boxes and interactive tote marker
• Move task config to jsk_arc2017_baxter
• Rename config collect_data -> collect_data_in_shelf
• add files for data collection
• Remove no need merge_depth_images.py
• add depth image merging nodes
• Deploy FCN32s trained on Dataset=v2, config=003
• add shelf_marker for stow_task
• add stow_json_generator and sample_stow_task json
• fix typo in pick_json_generator
• Add python module: get_object_names (#2132)
  • Add python module: get_object_names
  • Fix for flake8
  • Fix typo
• Make label_names.yml as just a name list
• Add log summarization script
• Improve logging in training script
• add easy picking task json files for mayfes demo
• Contributors: Kentaro Wada, Shingo Kitagawa, Shun Hasegawa, Yuto Uchimi

3.0.3 (2017-05-18)

3.0.2 (2017-05-18)

• Fix missing build depend on jsk_data
  • because install_data.py is run in Cmake
• Contributors: Kentaro Wada

3.0.1 (2017-05-16)

• Fix missing dependency on jsk_data
• fix typo in WorkOrderPublisher
• sort cardboard by box size and give ABC name
• Contributors: Kentaro Wada, Shingo Kitagawa

3.0.0 (2017-05-08)

• Fix style of nodes in roslaunch files
• Add sample for work_order_publisher
• Fix name of sample_set_location_in_rosparam
• Fix for move of data/objects -> config/objects
• Don't use ROS in training script
• add sample launch for set_location_in_rosparam
• print stdout in set_location_in_rosparam
• fix typo in set_location_in_rosparam
• remove unused package and sort alphabetically
• use label_names.yaml instead of objects.txt
• set myself as a author
• update json generator script
• mv pick_work_order_server -> work_order_publisher
• replace publish_shelf_bin_bbox to existing node
• remove unnecessary lines in CMakeLists
• move json -> data/json
• switch cardboard place cardboard a: left upper cardboard b: left lower cardboard c: right
• add abandon items for work_order_server
• fix typo in package.xml in jsk_arc2017_common
• update shelf_bin position config
• set cardboard id as A,B,C in work_order
• add pick_work_order_server test
• fix typo in arc2017 json item_location_file.json
• add myself as a maintainer
• update CMakelists.txt and package.xml for roseus
• add set_location_in_rosparam node
• format bin_name as capital alphabet
• update pick_work_order_server for new json format
• update json generator and sample in correct format
• add example json and box size config
• add pick_work_order_server for arc2017
• introduce new WorkOrder&WorkOrderArray msg
• add sample_pick.json and json generate script
• add setup_for_pick.launch for arc2017
• add shelf_interactive_marker.yaml
• add publish_shelf_bin_bbox for new shelf
• Add python-serial to run_depends
• Fix typo
• Read weight data from AND scale
  • new file: and_scale_rosserial.py
• Ignore AR20170331
• Update model file with stacking data augmentation
• Add data augmentation method with stacking
• Update api of torchfcn
• Improve imgaug
• Simplify config
• Update data with AR_20170331 dataset
• Add link to wiki
• Neat config & log handling
• Add ROS sample of FCNObjectSegmentation
• Add sample data of JSKV1 dataset
• Fix path of data
• Change path of JSKV1
• Add option to skip dataset with stamp
• Show datetime in annotation
• Improve view_jsk_v1
  • p for back
  • show timestamp
• Training experiments
• Update config
• Check label.npz existence
• Sort dirs for annotation
• Fix locking
• Show stamp_dir
• Lock for parallel annotation
• Augument image using imgaug
• Fix data field name
• 002_fcn32s_dataset_v1.yaml
• Fix for flake8
• Add requirements.txt
• Training script of FCN32s
• Add dataset class for JSKARC2017From16
• Add script to convert JSKAPC2016 to ARC2017
• Split dataset for train and valid
• Remove underscore for consistent names
• Add dataset.py
• Neat visualization of dataset
  • Show size of All and Annotated
  • Show label names
• Script to view dataset before/after annotated
• Update data using Training_items_20170320_fixed.zip
• Save data with compression
• Save label as npz file with compression
• Smaller size of object list
• Annotation script for JSK_V1 dataset
• Add script to list objects
• Visualize object list
• Parse AR_20170224 dataset
• Contributors: Kentaro Wada, Shingo Kitagawa