Amshaker / unetr_plus_plus

[IEEE TMI-2024] UNETR++: Delving into Efficient and Accurate 3D Medical Image Segmentation
Apache License 2.0
334 stars 30 forks source link

Dear author: may I ask why there is a big gap between the final results of my training and those in your paper? #60

Open liaochuanlin opened 10 months ago

liaochuanlin commented 10 months ago

"mean": { "0": { "Accuracy": 0.9981411858030146, "Dice": 0.9990577228352164, "False Discovery Rate": 0.0011644510163759305, "False Negative Rate": 0.0007191647780758607, "False Omission Rate": 0.06896922150842583, "False Positive Rate": 0.10025112646053021, "Jaccard": 0.9981179285963557, "Negative Predictive Value": 0.931030778491574, "Precision": 0.998835548983624, "Recall": 0.999280835221924, "Total Positives Reference": 8818659.328767123, "Total Positives Test": 8822584.849315068, "True Negative Rate": 0.8997488735394698 }, "1": { "Accuracy": 0.9974477693695682, "Dice": 0.806632188989613, "False Discovery Rate": 0.18348483329748716, "False Negative Rate": 0.18216677706851542, "False Omission Rate": 0.001330586076111128, "False Positive Rate": 0.0012451983567199761, "Jaccard": 0.689636518668635, "Negative Predictive Value": 0.998669413923889, "Precision": 0.8165151667025129, "Recall": 0.8178332229314845, "Total Positives Reference": 71735.91780821918, "Total Positives Test": 70985.54794520549, "True Negative Rate": 0.9987548016432801 }, "2": { "Accuracy": 0.9985986292163795, "Dice": 0.5921522768368525, "False Discovery Rate": 0.2734049053331573, "False Negative Rate": 0.4188343056970882, "False Omission Rate": 0.0009349991583760044, "False Positive Rate": 0.000470230968993352, "Jaccard": 0.4609482615097039, "Negative Predictive Value": 0.9990650008416241, "Precision": 0.7265950946668426, "Recall": 0.5811656943029119, "Total Positives Reference": 21344.64383561644, "Total Positives Test": 17205.178082191782, "True Negative Rate": 0.9995297690310067 }, "3": { "Accuracy": 0.9995000199464822, "Dice": 0.7779489448052025, "False Discovery Rate": 0.17003739413404176, "False Negative Rate": 0.21365629459137245, "False Omission Rate": 0.00019648618697882313, "False Positive Rate": 0.0003048707505856154, "Jaccard": 0.6751649761571616, "Negative Predictive Value": 0.9998035138130211, "Precision": 0.8299626058659582, "Recall": 0.7863437054086276, "Total Positives Reference": 16260.109589041096, "Total Positives Test": 17224.424657534248, "True Negative Rate": 0.9996951292494143 } } },

Amshaker commented 6 months ago

Please clarify the question. This is on which benchmark and within which folder?