yerang823/keras-retinanet

Result

Architecture

Usage

Preprocess

python train_test_split_txt.py

• When using k-fold cross validation, split annotation txt files.

Train

python keras_retinanet/bin/train.py --steps=1000 --workers=0 csv data/txt/train.csv data/txt/class_name.csv

Convert model

python keras_retinanet/bin/convert_model.py ./snapshots0/resnet50_csv_50.h5 ./final_model/resnet50_final.h5

Predict

python ResNet50RetinaNet.py

• Save result of images and txt files.

Evaluation

python evaluate_classification.py