Step CV-Pipeline: model_train

This CV-Pipeline component is intended for configuring model training and starting model training

Input data for step CV-Pipeline: model_train

• coco_train_dataset
  Training dataset
• coco_eval_dataset
  Validation dataset
• yolox_pth_pretrain_weights
  Predtraining weights

The output of this step CV-Pipeline is

• obj_detect_inference_files
  Saved weights of the trained model (weights of the last epoch and with the best achieved metrics), configuration files

How to run a step CV-Pipeline: model_train

Create a directory for the project (or use an existing one)

mkdir -p obj_detect_rest
cd obj_detect_rest

clone the repository: model_train

git clone --recurse-submodules https://github.com/4-DS/obj_detect_rest-model_train.git {dir_for_model_train}
cd {dir_for_model_train}

run step CV-Pipeline:model_train

python step.dev.py

or

step.prod.py