unicef/Sudan-school-mapping-AI_models

Giga_Omdena project # UNICEF_Talent ## Content 1. Project directories structure 2. Model files * 2.1 Tensorflow models * 2.2 YOLTv5 * 2.2.1 How it works * 2.2.2 How to use 3. Pipeline * 3.1 How pipeline works ## Project directories structure ``` data └─── models - saved models and weights └─── temp - temporary folder containing images slices, created only during debugging src - contains the entire source code └─── stages - pipeline stages scripts │ └─── preprocessing.py - preprocessing tasks │ └─── prediction.py - prediction running tasks │ └─── postprocessing.py - postprocessing tasks └─── tf - source code related to TensorFlow solution │ └─── models - TensorFlow models │ │ └─── common.py - base class implementation inherited by all TensorFlow models │ │ └─── ensemble.py - an ensemble model implementation that combines results from object classifier and object localizer │ │ └─── object_localizer.py - object localization model │ │ └─── object_classifier.py - object classification model │ │ └─── loss_learning_rate_scheduler.py - adaptive learning rate scheduler │ │ └─── loss_utils.py - custom loss functions │ │ └─── model_utils.py - utilities for working with models │ │ └─── patch_encoder.py - patches encoder for the ViT model │ │ └─── patches.py - creates patches for the ViT model │ │ └─── resnet.py - common functionalities inherited by all ResNet models │ │ └─── resnet101.py - 101-layer ResNet model │ │ └─── resnet18.py - custom 18-layers ResNet model │ │ └─── resnet5 …