RSANet_PCam v0.1.0 Pretrained Weights (Seed2)

IRSANet_PCam v0.1.0 Pretrained Weights (Seed2)

Model Details

Field	Value
Model	IRSANet_PCam
Parameters	347,775
Modules	124

Weight Details

Field	Value
Experiment	Validation
Run	Run4
Seed	2
Fold	single
Epoch	17
Checkpoint	bestAP

Evaluation Details

Field	Value
Dataset	PCam
Test samples	32,768
Evaluated	2026-06-04 07:49
TTA	False

Key Metrics (Threshold = 0.441)

Metric	Test	Val
AUC (ROC)	0.967	0.969
Average Precision	0.970	0.970
F1 Score	0.904	—
Balanced Accuracy	0.906	—
Recall (Sensitivity)	0.883	—

Usage

Reconstruction

from irsanet import load_checkpoint

model, ckpt = load_checkpoint("IRSANet_PCam_seed2_bestAP.pt")

Evaluation

from irsanet import evaluate_model
from irsanet_pcam import build_datasets

_,_,test_loader,_,_ = build_datasets([],
    eval_mode  = True,
    seed       = 2,
)

results = evaluate_model(
    weights     = "IRSANet_PCam_seed2_bestAP.pt",
    test_loader = test_loader,
)

print(results["test_results"])

Including a full visual dashboard:

from irsanet.plotting import plot_eval_panel
from irsanet_pcam import build_datasets

_,_,test_loader,_,_ = build_datasets([],
    eval_mode  = True,
    seed       = 2,
)

results = plot_eval_panel(
    weights     = "IRSANet_PCam_seed2_bestAP.pt",
    test_loader = test_loader,
)

print(results["test_results"])