This repository has been archived by the owner on Dec 1, 2021. It is now read-only.
Releases: blue-oil/blueoil
Releases · blue-oil/blueoil
v0.30.0
Bug Fixes and Refactoring
- replace sprintf for ostringstream(#1192) (#1199) by @lm-kajihara
- replace network::run() argument types to void* (#1223) by @lm-kajihara
- Fix PascalVOC dataset loader to accept float bndbox (#1228) by @oatawa1
Documentation
- move keep_me.md(#937) (#1200) by @lm-kajihara
v0.29.0
Features and Refactoring
- Replace EasyDict to SmartDict (#1220) by @iizukak
- Remove output_event.py (#1219) by @iizukak
- Consider that 'Reshape' won't preserve quantization (#1217) by @primenumber
- Support evaluation with GCS I/O (#1206) by @hadusam
- Avoid ValueError for TFDS with BBox on the last pixel (#1209) by @hadusam
- rewrite error comparing (#1214) @lm-kajihara
- Adding
--gpu={gpu index}flag to docker option in tutorials. (#1212) by @ananno - Enable to save pretrain_vars.txt to GCS (#1197) by @hadusam
- change python to python3 in usage docs (#1202) by @joelN123
- Add blueoil.io module for I/O with GCS support (#1204) by @hadusam
- change the option name in export documentation to "image" (#1203) by @joelN123
- Update converter test to PY3 and drop py2 functions (#1154) by @kchygoe
- Remove xenial from test pipelines (#1198) by @kchygoe
Bug Fixes
- Fix confitions not to use tfds_pre_processor and tfds_augumentor (#1207) by @hadusam
- fix : Change time.clock to time.perf_counter (#1191) by @lm-kasuga
Documentation
v0.28.0
Features / Refactoring
- Add unittest CI environment for inference scripts (output_template) #1155 by @hadusam
- Add unit test for PASCALVOC Dataset Loader classes (#1166) by @oatawa1
- Add dataset loader for FER2013 format (#1172) by @Suikaba
- Change offsetVolume function (#1165) by @tk26eng
- Enable coverage report in unit tests (#1175) by @tfujiwar
- Add export network trainable variables (PRETRAIN_VARS) to experiment dir #1179 by @oatawa1
- Update the link to the latest OS image (#1182) by @Suikaba
- Add explanations about serial communication (#1183) by @lm-kasuga
- Remove the unnecessary subset argument (#1186) by @Suikaba
- Add a config file for FER2013 (#1185) @Suikaba
- Lookup without DivideBy255 (#1161) by @primenumber
- Use g++-8 to keep consistency in each architecture (#1190) by @hadusam
- Fix mypy errors: converter/core/operators.py (#1146) by @tfujiwar
- Python3 support of motion_jpeg_server_from_camera.py (#1156) by @hadusam
Bug Fixes
- Fix argument order for Resize function (#1159) by @tk26eng
- Fix absl-py version to 0.8.1 (#1168) by @oatawa1
- Fix shapes and available_subset of FER2013 loader (#1177) by @Suikaba
- Fix an error of runtime with no bounding box of object detection (#1163) by @tk26eng
- Fix MAX_STEPS = 0 error by add suggestion and exit (#1195) by @oatawa1
Documentation
- Fix missing link in document (#1187) by @oatawa1
- Fix document typo (#1180) by @lm-kasuga
v0.27.0
Features and Refactoring
- Insert Cast operator automatically (#1119) by @primenumber
- Add minimum package installation and network settings to rootfs setup scripts (#1127) by @hadusam
- Add blank line between copyright header and code #1133 by @oatawa1
- Add DE10-nano bionic test env (#1130) by @kchygoe
- Add gitattribute to exclude files from archive (#1132) by @kchygoe
- Remove hw directory and replace it in GCS (#1131) by @kchygoe
- Switch main test DE10-Nano OS to bionic (#1138) by @kchygoe
- capitalize help messages (#1137) by @joelN123
- Remove pandas dependency (#1072) by @lm-lily
- CI: Add mypy test (#1141) by @tfujiwar
- Upgrade OS version to bionic for Docker image (#1139) by @kchygoe
- Fix comments typo in base.py (#1135) by @oatawa1
- Enhance the dataset class to accept the dataset in PASCALVOC format with configuration. #1140 by @oatawa1
- Update Raspberry-Pi CI env (#1147) by @kchygoe
- Update dependencies (#1151) by @kchygoe
- Remove make_os directory (#1153) by @hadusam
v0.26.0
Features / Refactoring
- remove delta-mark files (#1086) by @lm-lily
- Profiler for training (memory and time per operation) of one train step (#1111) by @joelN123
- One operator has at most one output (#1113) by @primenumber
- Remove unused functions: write_to_file (#1116) by @primenumber
- Divide global.h into global.h and parameters.h (#1117) by @primenumber
- Rename depth to channels (#1118) by @primenumber
- Add turbo color map (#1055) by @kalpitthakkar-lm
- Change a default bit of profiling (#1125) by @hadusam
Bug Fixes
- Fix limitation (#1114) by @primenumber
- FIx to set correct gs:// path (#1122) by @iizukak
Documentation
v0.25.0
Features / Refactoring
- tf_upgrade_v2 for networks/classification/quantize_example.py (#1080) by @ytfksw
- remove util functions that are no longer used (#1098) by @joelN123
- Support FPGA with aarch64 (#1063) by @tk26eng
- Fix license header consistency (#1093) by @oatawa1
- Check stride and dilation >0 (#1100) by @primenumber
- Report maximum diff (#1109) by @primenumber
- Implemented easydict. (#1094) by @odoku
- Update tutorials to use CPU in converting step (#1099) @tk26eng
Documentation
v0.24.0
Features / Refactoring
- Remove unused inheritance (#1054) by @primenumber
- Added "data_dir" and "output_dir" options to the command. (#1049) by @odoku
- Support import of Sub operator (#941) by @nlpng
- Kernel Initializer issue (#364) by @ananno
- Update default config for BDD100K dataset (#754) by @ananno
- Added resample option to resize processors. (#1060) by @odoku
- Added test_image option to convert command. (#1056) by @odoku
- Processing has been optimized and speed has been improved. (#1068) by @odoku
- Refactor cpp code (#1071) by @lm-noshiro
- Change colormap scope (#1034) by @kchygoe
- Use time.time/time.perf_counter instead of time.clock for demo scripts by @hadusam
- TF2.0 support changes #1073 #1074 #1075 #1067 #1077 #1064 #1069 by @ytfksw
Bug Fixes
- Missing outputs of input placeholder (#1045) by @nlpng
- Fix checking pad (#1085) by @primenumber
- Fix bug without cache (#1048) by @tk26eng
Documentation
v0.23.0
Features / Refactoring
- Remove an argument
typefrom all__getitem__method (#973) by @tfujiwar - Refactor for consistency in blueoil/networks directory (#974) by @oatawa1
- Add a simple network as a supplementary source code for how to quantize document. (#966) by @oatawa1
- Remove SSDRandomCrop augmentation from classification task tutorial. (#978) by @oatawa1
- Add init.py to the missing locations. (#994) by @oatawa1
- Change condition to show an warning from generate_project.py (#982) by @tk26eng
- Remove some code from time measurement (#986) by @tk26eng
- Include converter in pep8 coverage (#988) by @tfujiwar
- Create auto generate Python API document for
blueoil/directory. (#995) by @oatawa1 - support TFDS format for segmentation (#1005) by @hadusam
- Add table support in markdown (.md) file (#1004) by @oatawa1
- change available_subsets as a class variable (#1008) by @hadusam
- Move global_step to network (#999) by @tfujiwar
- Clean Up Makefile of Runtime (#1013) by @tk26eng
- Remove "network" and "dataset" args from train end evaluate command (#1012) by @tfujiwar
- Remove _init_flag from environment module (#1009) by @tfujiwar
- Remove unused global variables in environment module (#1010) by @tfujiwar
- Remove first result from measurement (#1019) by @tk26eng
- Update python patch version and setup.cfg (#1024) by @kchygoe
- Install liblzma-dev in docker image (#1023) by @tfujiwar
- Replace deprecated
np.asscalar()withnp.ndarray.item()(#1022) by @tfujiwar - Apply tf_upgrade_v2 for quantizations (#915) by @ytfksw
- Remove deprecated properties from dataset class (#1025) by @yd8534976
- Optimize binary operation (#1027) by @primenumber
- Add PerImageStandardization compatibility in (C++ inference) runtime #1031 by @oatawa1
- Add missing C++ runtime unit test (#1037) by @oatawa1
- Modify unused part (#1040) by @tk26eng
- Delete test directory (#1041) by @tk26eng
Bug Fixes
- Fixed pep8 errors "F405" in converter (#981) by @tfujiwar
- Fixed pep8 errors "F841" in converter (#984) by @tfujiwar
- Fixed pep8 errors "F401" in converter (#983) by @tfujiwar
- Fix pep8 errors about formatting (#985) by @tfujiwar
- Fix Blueoil init in the case of no validation dataset(#976) by @oatawa1
- Fixed pep8 errors about exceptions (#987) by @tfujiwar
- Remove validation dataset path from object detection tutorial document (#977) by @oatawa1
- Remove error from make html doc and update converter doc using Sphinx-apidoc (#989) by @oatawa1
- Fix a stability about multiple measurements (#1001) by @tk26eng
Documentation
- How to quantize document (#955) by @oatawa1
- Move support ops to the end of converter document part (#1016) by @oatawa1
- Update documents to follow latest Makefile (#1015) by @tk26eng
- Add specification to blueoil document index. (#1014) by @oatawa1
- Add Network specification and update network code structure. (#1000) by @oatawa1
- Fix image_keypoint_detection tutorial document typo (#1032) by @oatawa1
- Update (C++ inference) runtime documents (#1030) by @oatawa1
- Add documentation about png input in C++ runtime (#1038) by @oatawa1
v0.22.0
Major features
Features / Refactoring
- Apply tf_upgrade_v2 (#914) by @ytfksw
- Apply tf_upgrade_v2 (#916) by @ytfksw
- Add Comment (#925) by @primenumber
- Add make commands (#922) by @kchygoe
- Include the tests directory to flake8 targets (#932) by @tfujiwar
- Update flake8 and fixed lint (#923) by @kchygoe
- Add time measurement flag to all elf binaries in Makefile (#928) by @tk26eng
- Raise a FileNotFoundError when any config file does not exists (#943) by @tfujiwar
- Report error and abort if channel size unmatch in ConcatOnDepth (#950) by @primenumber
- Rename QTZ_linear_mid_tread_half (#670) by @Joeper214
- Improve readability of the predict script (#942) by @tfujiwar
- Add divide 255 pre processor test (#845) by @Joeper214
- Remove a dataset subset type "train_validation_saving" (#948) by @tfujiwar
- Remove graph output logic from training loop (#947) by @tfujiwar
- Delete cmake dependency (#936) by @tk26eng
- Update tfds version to 1.3.0 (#956) by @tfujiwar
- Remove a memset function (#957) by @tk26eng
- Set yaml default Loader (#958) by @Joeper214
- Add device inference test to e2e tests (#872) by @hadusam
- Remove metrics placeholder from training loop (#959) by @tfujiwar
- unpacking debug file for quantized tensor (#934) by @primenumber
- Add _save_json unit test for profile model (#952) by @Joeper214
- Avoid to use static constexpr member in C++14 (#951) by @primenumber
- Apply tfds pre processor (#846) by @kurarrr
- Update version of Pillow library to fix security problem (#970) by @tk26eng
Bug Fixes
- Fix debug (#918) by @primenumber
- Fix very small comment typo (#926) by @oatawa1
- Fix run.py time measurement log output (#930) by @kchygoe
- Fix threshold skipping if flg = -1 (#917) by @primenumber
- Fix comparison order in generationg thresholds (#912) by @primenumber
- Fix checkpoint loader to use GFile (#938) by @kchygoe
- Fix comment typo in inin.py in blueoil/blueoil/quantizations (#944) by @oatawa1
- Fix softmax to avoid overflow (#945) by @primenumber
- Fix typo on darknet.py and lm_fyolo.py (#963) by @oatawa1
- Fix condition for threshold-skipping (#919) by @primenumber
- Fix arg name in comment quantize to quantizer (#967) by @oatawa1
Documentation
v0.21.0
Features / Refactoring
- tf_upgrade_v2 for cmd (#875) by @ytfksw
- tf_upgrade_v2 for dataset (#894) by @ytfksw
- tf_upgrade_v2 for configs (#899) by @ytfksw
- convert tf2 for blueoil/generate_lmnet_config.py (#902) by @ytfksw
- Add blueoil.datasets as module (#887) by @kchygoe
- Migrate from whaaaaat to inquirer (#882) by @oatawa1
- Fix LMBiSeNet inference (.elf) test error on x86 and DE10-Nano (#886) by @oatawa1
- Experimental support for 5x5 convolution on CPU (#683) by @primenumber
- Remove PYTHONPATH (#907) by @primenumber
- Update dependency packages (#910) by @kchygoe
- Include types.h instead of global.h (#891) by @primenumber
Bug Fixes
- Fix type annotation (#890) by @primenumber
- Fix comparator for decreasing function (#888) by @nlpng
- Fix output image results in Semantic Segmentation task from run.py (#889) by @oatawa1
- Fix code generator typo (#900) by @Joeper214
Documentation
- Remove unused DLK readme files (#878) by @Joeper214
- Move lmnet/README.md to docs/usage and split it into each subcommand #893 by @tfujiwar
- Update DLK -> Converter docs (#896) by @Joeper214
- Update blueoil tutorial document to match blueoil r 0.20 (#903) by @oatawa1