How to interpret stardist.matching results?
fangpingmu opened this issue · comments
model = StarDist2D.from_pretrained('2D_versatile_he')
img_num = 0
img = X[img_num]
img = normalize(img, 1,99.8, axis=axis_norm)
labels, details = model.predict_instances(img) #,prob_thresh=0.4)
metrics = matching(Y[img_num], labels)
print(metrics)
The image is the first MoNuSeg testing image. I don't perform any normalization.
Matching(criterion='iou', thresh=0.5, fp=514, tp=57, fn=134, precision=0.09982486865148861, recall=0.29842931937172773, accuracy=0.08085106382978724, f1=0.14960629921259844, n_true=191, n_pred=571, mean_true_score=0.21061019498016198, mean_matched_score=0.7057288989686129, panoptic_quality=0.10558148882207595)
I cannot interpret the matching results. Precision, recall, accuracy, f1 are so low. When I view the prediction results (left is grand truth and right is prediction mask), the prediction looks good
.Dear @fangpingmu, I'm sorry but this is a general question and not a bug report. Please ask this question at the forum, as instructed.