How shahrukhqasim/ties-2.0 works
TIES-2.0 extracts table structure from images using graph-neural-network models. The workflow is configured dataset input, iterative training or inference, then inference-output analysis and plots. The README documents training, inference, TFRecord datasets, and model variants; sampled source confirms the iterator/model/data and output-writing relationships. Some model internals and plotting/feedback implementations are indexed but not excerpted, so their undocumented connections are omitted.
Main entry points: model_factory.py Model factory · basic_model.py Table graph model · table_adjacency_parsing_iterator.py Train and test iterator · table_adjacency_parsing.py CLI entrypoint
Components 20
Run workflow
- Run configuration
configs/config.ini.example - CLI entrypoint
python/bin/iterate/table_adjacency_parsing.py - Train and test iterator
python/iterators/table_adjacency_parsing_iterator.py - Configuration manager
python/libs/configuration_manager.py
Table data
- TFRecord dataset
- Image and word reader
python/readers/image_words_reader.py
Model architecture
- Model factory
python/models/model_factory.py - Table graph model
python/models/basic_model.py - Convolution segments
python/models/fast_conv_segment.py - Basic convolution segment
python/models/conv_segment.py - DGCNN segment
python/models/dgcnn_segment.py - GarNet segment
python/models/garnet_segment.py - FCNN segment
python/models/fcnn_segment.py - GravNet segment
python/models/gravnet_segment.py - Graph operations
python/ops/ties.py
Results and analysis
- Inference output files
python/libs/inference_output_streamer.py - Inference analyser
python/bin/analyse/table_adjacency_inference_analyser.py - Result plots
python/libs/plots.py - Visual feedback
python/libs/visual_feedback_generator.py
Other components
- Researcher
Connections 22
- Researcher → CLI entrypoint: runs
README.md - CLI entrypoint → Configuration manager: initializes
python/bin/iterate/table_adjacency_parsing.py - Run configuration → CLI entrypoint: configures
README.md - Train and test iterator → Configuration manager: reads settings
python/iterators/table_adjacency_parsing_iterator.py - CLI entrypoint → Train and test iterator: dispatches
python/bin/iterate/table_adjacency_parsing.py - Train and test iterator → Model factory: requests model
python/iterators/table_adjacency_parsing_iterator.py - Model factory → Table graph model: creates
python/models/model_factory.py - Model factory → Convolution segments: selects
python/models/model_factory.py - Model factory → DGCNN segment: selects
python/models/model_factory.py - Model factory → GarNet segment: selects
python/models/model_factory.py - Model factory → FCNN segment: selects
python/models/model_factory.py - Model factory → GravNet segment: selects
python/models/model_factory.py - Model factory → Basic convolution segment: imports variant
python/models/model_factory.py - Train and test iterator → Table graph model: runs iterations
python/iterators/table_adjacency_parsing_iterator.py - TFRecord dataset → Image and word reader: provides records
python/readers/image_words_reader.py - Table graph model → Image and word reader: loads batches
python/models/basic_model.py - Table graph model → Graph operations: uses operations
python/models/basic_model.py - Table graph model → Inference output files: writes samples
python/models/basic_model.py - Table graph model → Visual feedback: generates feedback
python/models/basic_model.py - Inference output files → Inference analyser: supplies predictions
python/bin/analyse/table_adjacency_inference_analyser.py - Inference analyser → Configuration manager: loads settings
python/bin/analyse/table_adjacency_inference_analyser.py - Inference analyser → Result plots: plots results
python/bin/analyse/table_adjacency_inference_analyser.py