IRSANet_PathMNIST v0.1.0 Pretrained Weights (Seed3)

IRSANet_PathMNIST v0.1.0 Pretrained Weights (Seed3)

Model Details

Field	Value
Model	IRSANet_PathMNIST
Parameters	351,623
Modules	124

Weight Details

Field	Value
Experiment	Validation
Run	Run4
Seed	3
Fold	single
Epoch	99
Checkpoint	bestAP

Evaluation Details

Field	Value
Dataset	PathMNIST
Test samples	7,180
Evaluated	2026-06-05 10:52
TTA	False
Classes	9

Key Metrics (Macro)

Metric	Test	Val
AUC (ROC, OVR)	0.997	1.000
Average Precision	0.974	0.999
F1 Score	0.951	0.993
Balanced Accuracy	0.952	0.993

Usage

Reconstruction

from irsanet import load_checkpoint

model, ckpt = load_checkpoint("IRSANet_PathMNIST_seed3_bestAP.pt")

Evaluation

from irsanet import evaluate_model
from irsanet_pathmnist import build_datasets

_,_,test_loader,_,_ = build_datasets([],
    eval_mode  = True,
    seed       = 3,
)

results = evaluate_model(
    weights     = "IRSANet_PathMNIST_seed3_bestAP.pt",
    test_loader = test_loader,
)

print(results["test_results"])

Including a full visual dashboard:

from irsanet.plotting import plot_eval_panel_multi
from irsanet_pathmnist import build_datasets

_,_,test_loader,_,_ = build_datasets([],
    eval_mode  = True,
    seed       = 3,
)

results = plot_eval_panel_multi(
    weights     = "IRSANet_PathMNIST_seed3_bestAP.pt",
    test_loader = test_loader,
)

print(results["test_results"])