Gs/improvements from stvw

[GwenaelleSa]

• Make it possible to resume training
• Add data augmentation for color thank to a modified dataset mapper
• Remove duplicated passing of the model

[GwenaelleSa]

I updated the branch with main, which took care of your comment @acerioni
About the learning rate, I changed it because it did not make sense to me that the best model was reached after 500 steps. It's just as long as the warm-up period of the training. Even if the total loss looks pretty, it's still probably not necessary for it to be as steep.

Initial learning curves

Learning curves with base Lr 0.0005

[stdl-s-sonarqube]

Analysis Details

2 Issues

• 0 Bugs
• 0 Vulnerabilities
• 2 Code Smells

Coverage and Duplications

• No coverage information (0.00% Estimated after merge)
• 0.00% Duplicated Code (0.00% Estimated after merge)

Project ID: swiss-territorial-data-lab_object-detector_AYZ4zWIzr7JdaaSXwexc

View in SonarQube