Fig. 11From: Contrastive self-supervised representation learning framework for metal surface defect detectionValidation loss curve of ConvNet: baseline, pretraining, and downstream/finetuningBack to article page