| license: apache-2.0 | |
| library_name: structflo-cser | |
| tags: | |
| - object-detection | |
| - chemistry | |
| - document-analysis | |
| # cser-lps — structflo-cser weights | |
| Learned Pair Scorer — a ~557K-parameter CNN that scores (structure, label) box pairs from geometric features and visual crops (our own architecture, plain torch state_dict). | |
| | field | value | | |
| |---|---| | |
| | latest version | `v0.3` (HF tag `weights-v0.3`) | | |
| | file | `best.pt` | | |
| | sha256 | `3c5914e8798e9195820eda92c87824c53fb05a81fff690828388416331eaa531` | | |
| | requires | `structflo-cser>=0.4.0,<1.0.0` | | |
| ## Usage | |
| ```python | |
| from structflo.cser.pipeline import ChemPipeline | |
| pipeline = ChemPipeline() # resolves the latest registered weights | |
| pairs = pipeline.process("page.png") | |
| ``` | |
| Pin a version with `ChemPipeline(weights="v0.3")`. | |
| ## Licence and provenance | |
| Weights: Apache-2.0. Training data: synthetic pages rendered with RDKit from ChEMBL | |
| (CC-BY-SA-3.0) SMILES plus an internal annotated corpus (not released). See the package's | |
| `THIRD_PARTY_NOTICES.md` for upstream attributions. | |