Table 3 Performance of the fine-tuned models. Each pre-trained model is trained seven times. From Train1 to Train7, new training images are added one at a time in the order shown in Table 1. For train7, two performance metrics are shown: mAP and ARI, where ARI is the average Adjusted Rand Index over seven test images

Cellpose

cyto

0.518

0.453

0.607

0.633

0.657

0.668

0.691

0.694

0.815

cyto2

0.478

0.460

0.616

0.632

0.652

0.661

0.689

0.689

0.817

tissuenet

0.001

0.471

0.598

0.604

0.630

0.647

0.671

0.676

0.813

livecell

0.211

0.420

0.597

0.591

0.635

0.648

0.665

0.673

0.812

none

0.293

0.510

0.510

0.576

0.589

0.616

0.615

0.791

DeepCell

tn_nuclear

0.387

0.373

0.381

0.380

0.392

0.426

0.452

0.488

0.668

tn_cyto

0.140

0.277

0.349

0.337

0.357

0.355

0.423

0.428

0.648

MR-CNN

jacs

0.350

0.530

0.542

0.519

0.548

0.558

0.576

0.576

0.706