How nengdong96/etndnet works
Project type and purpose
• ETNDNet is a Python-based deep learning training and evaluation framework (no web UI or microservices).
• It implements a full model training pipeline: data loading & preprocessing, model definition, training & testing loops, evaluation metrics, and visualization.
File structure
• Top-level scripts:
• main.py – entry point for training/testing
• run.sh – shell wrapper to invoke main.py with args
• core/ – orchestrates training/testing logic
• base.py – base classes for trainers/testers
• train.py/test.py – train and test workflows
• data_loader/ – data ingestion & preprocessing
• loader.py/dataset.py – dataset definitions & IO
• preprocessing.py/data_process.py – transforms & augmentation
• sampler.py – custom sampling strategies
• network/ – neural network components
• model.py – top-level model class
• res50ibnabnneck.py – specific ResNet-based architecture
• feature_map_processing.py/gem_pool.py – feature pooling operations
• tools/ – supporting utilities
• loss.py – loss function definitions
• evaluation/ – classification & re-identification metrics
• metric.py/meter.py – running metrics & statistics
• logger.py/utils.py – logging & helper functions
README for additional insights
• README only states “# ETNDNet,” so all architecture details must be inferred from folder names and common DL practices.