Releases: Kohulan/DECIMER-Image_Transformer
Releases · Kohulan/DECIMER-Image_Transformer
v2.6.0
v2.5.0
v2.4.0
2.4.0 (2023-08-30)
Features
Bug Fixes
- add pyyaml to requirements.txt (d572ded)
- confidence extraction (6c95fee)
- efficientnet imports (70080f6)
- efficientNet Model params (7057fab)
- Import modules according to TensorFlow versions (65600ff)
- linter issue of unused import (5f607e4)
- linting issues (40cffed)
- model size to check for download (b66bca6)
- read the docs theme (f6732e7)
- remove duplicate efficientnet imports (3896425)
- revert to original way of importing (7af54c9)
- update EfficientNet import issue (758a0f4)
- use checkpoints properly for predictions #65 and #66 (f47ac99)
Reverts
- original implementation (e1b946a)
v2.3.1
2.3.1 (2023-07-04)
Features
- Add new theme for docs (4d69afd)
- add release-please (bbfdc80)
- Create Bug_Report.yml (ffb8540)
- Create CODE_OF_CONDUCT.md (605fdb4)
- Create ISSUES.md (82ebb33)
- New model and preprocessing (782c5b9)
- Update config file and refactor code (c46c852)
- Update repack to fit to 512 x 512 (960904b)
Bug Fixes
- check HEIC from path #44 (40bb299)
- check model path (aacbe3a)
- delete model zipfile (bab2193)
- Docs issues for building (e715966)
- docs requirements.txt (ced96a3)
- Don't download model if present (57e809f)
- Fixed Benchmark/README example (aaf4db2)
- flake8 errors (99eb533)
- Import efficient-net properly (16b8e63)
- import pyheif (6ecb4e4)
- linter issues (839c0a2)
- modify branch (6a17ee3)
- Prediction postprocessing (063a50f)
- print original error (504b394)
- pyheif isuue for windows using pillow-heif (49bae11)
- raise exception (b2dad4d)
- Readthedocs theme (26c3ba0)
- Removes previous models (fb9b858)
- revert TF version (0f99e12)
- set TensorFlow version to 2.10.0 (1221cad)
- start release version (964599d)
- Test assertion (aaf0a55)
- Test expected results (54dbbc8)
- typo Update Bug_Report.yml (20efde3)
- Update Bug_Report.yml (8a01bc4)
- Update contrast setting (f496307)
- Update DOI (71acf86)
- Update model links (d95a26a)
- Update TensorFlow support for MacOS (d4eb411)
- update tensorflow version (d0967b5)
- Update TensorFlow, fix: Tests (1d2b583)
- Update to load functions from installation #38 (960ffc6)
- Update tox.ini (7615d69)
- Update TRC Logo (3958dc6)
- Verion bump (c55769d)
- Verion bump setup.py (61049ef)
- version update (56b18ed)
DECIMER V2.3.0
- DECIMER V2.3.0 release
- Improved preprocessing pipeline
- New model released
- Current model understands explicit hydrogens as hydrogen isotopes
- Current model can interpret hand-drawn structures
- supports TF 2.12.0
DECIMER V2.2.0
- DECIMER can now handle images with higher resolution (512 x 512)
- New models are now available in Zenodo.
- The fine-tuned model is available for hand-drawn images.
- The latest models can now understand chemical structure depictions with R-Groups and predict SMILES
- The model was trained on images generated using RanDepict 1.1.3
Full Changelog: V2.1.0...V2.2.0
DECIMER Image Transformer Version 2.1.0
- DECIMER models have been updated in Zenodo.
- With the latest models being able to understand chemical structure depictions with R-Groups and predict SMILES.
- Images generated with RanDepict 1.0.9 were used to train the model
Full Changelog: V2.0.2...V2.1.0
DECIMER V2.0.2
- Minor updates
- Models moved to Zenodo instead of Google cloud buckets
Full Changelog: V2.0...V2.0.2
DECIMER V2.0 Release
DECIMER V2.0
- Features a EfficientNet-V2 Encoder+ Transformer Decoder
- Trained on whole data from PubChem
- Models trained on TPUs on Google Cloud
- PyPi version updated
- Removed multiple models and uses a single model for all tasks
- Higher accuracy than previous version
- Models optimized to work faster
- Web version of the prediction server is now available at https://decimer.ai by @OBrink
Contributors
OBrink
Assets 2
DECIMER Image Transformer Version 1.0.2
- Fixed decimer cli
- Added support for integrating decimer into python programs
- Fixed trained model download directories
- Code reformatted
- Minor bug fixes