当前位置： 首页 > java >正文

改进系列（12）：基于SAM交互式点提示的UNet腹部多脏器分割方法研究

java 2025/7/27 23:19:51

目录

1.摘要

2.引言

3.方法

3.1 网络架构

3.2 点提示机制

3.3 损失函数与优化

4.实验

4.1 数据集与实现细节

4.2 评价指标

4.3 结果与分析

5.讨论

6.结论

7.补充图

8.代码

1.摘要

本文提出了一种基于交互式点提示的UNet网络结构，用于腹部多脏器医学图像分割任务。该方法通过引入点提示机制，允许用户在推理阶段通过点击前景和背景区域提供交互式指导，显著提高了分割精度。我们在腹部多脏器MR数据集上进行了实验验证，结果表明该方法在Dice系数和IoU指标上分别达到了0.9358和0.8805的平均值。与传统全自动分割方法相比，本方法具有更高的灵活性和实用性，特别适用于医学图像分析中需要专家干预的复杂场景。

关键词：医学图像分割；UNet；交互式分割；点提示；深度学习

2.引言

医学图像分割是计算机辅助诊断系统中的关键技术，尤其在腹部多脏器分割任务中，精确的分割结果对于疾病诊断、手术规划和疗效评估具有重要意义。然而，由于腹部器官形状复杂、边界模糊且个体差异大，传统全自动分割方法往往难以达到临床要求的精度。

近年来，交互式分割方法逐渐受到关注，它将专家知识引入分割过程，通过简单交互操作指导算法获得更准确的结果。本文提出了一种改进的UNet架构，通过在输入通道中加入点提示信息，实现了基于用户点击的交互式分割。该方法在保持UNet高效特征提取能力的同时，充分利用了用户提供的空间先验信息，显著提升了分割性能。

原始标签：

3.方法

3.1 网络架构

本文采用改进的UNet作为基础网络结构。与传统UNet相比，主要做了以下改进：

  1. 输入通道扩展为4通道(原始3通道RGB图像+1通道点提示信息)

  2. 输出通道设置为2，使用softmax激活实现二分类

网络由编码路径、解码路径和瓶颈层组成。编码路径包含4个下采样块，每个块由两个3×3卷积、批归一化和ReLU激活组成，后接2×2最大池化。解码路径通过转置卷积上采样，并与编码路径相应尺度的特征图拼接，再经过两个3×3卷积。

3.2 点提示机制

点提示是本方法的核心创新点，其实现包含两个阶段：

训练阶段：从真实标注中随机采样前景点作为正提示(值为1)，当图像无前景时采样背景点作为负提示(值为-1)。这些点被编码为与图像尺寸相同的单通道矩阵，与原始图像拼接形成4通道输入。

推理阶段：用户通过GUI界面点击图像指定前景(左键)和背景(右键)，系统将这些点击位置转换为点提示通道，与图像一起输入网络获得分割结果。

3.3 损失函数与优化

采用交叉熵损失函数：

  • 其中N为像素数，C为类别数，y为真实标签，p为预测概率。

使用AdamW优化器，初始学习率设为0.001，采用余弦退火策略调整学习率：

  • 其中t时刻学习率，T为总epoch数。

4.实验

4.1 数据集与实现细节

实验使用腹部多脏器CT数据集，包含394+98例患者的扫描图像，按7:3划分为训练集和验证集。所有图像统一缩放到256×256像素，并进行归一化处理。

训练在NVIDIA RTX 4060 GPU上进行，batch size设为4，训练100个epoch。

4.2 评价指标

采用以下指标评估分割性能：

  • Dice相似系数(DSC)

  • 交并比(IoU)

  • 像素准确率(PA)

  • 精确率(Precision)

  • 召回率(Recall)

  • F1分数

4.3 结果与分析

下表展示了本方法在验证集上的平均性能指标：

评价指标本方法结果理想临床阈值
像素准确率(PA)0.9962>0.95
平均精确率0.9308>0.85
平均召回率0.9303>0.85
平均F1分数0.9306>0.85
平均Dice系数0.9306>0.85
平均IoU0.8702>0.75

下图展示了训练过程中的损失曲线和指标变化，可见模型在约50个epoch后趋于收敛。

下图为加入3个点提示后的交互式分割结果。可见交互式方法能有效修正自动分割的错误，特别是对于边界模糊区域。

5.讨论

本研究的创新点在于将点提示机制与UNet结合，实现了高效的交互式分割。与传统方法相比具有以下优势：

  1. 灵活性：允许专家根据实际需要调整分割结果

  2. 高效性：少量交互即可显著改善分割质量

  3. 实用性：GUI界面友好，便于临床使用

然而，该方法仍存在一些局限性：首先，点提示位置对结果影响较大，不合理的点选择可能导致次优分割；其次，当前实现仅支持单器官分割，同时分割多个器官需要多次交互。未来工作将探索多类别点提示和自适应点选择策略。

6.结论

本文提出了一种基于点提示的交互式UNet腹部多脏器分割方法。实验结果表明，该方法能有效利用用户提供的简单交互信息，显著提高分割精度。在临床应用中，这种方法可以平衡自动化效率和专家干预需求，为医学图像分析提供了一种实用解决方案。

7.补充图

 Namespace(base_size=(224, 224), batch_size=4, data_train='./data/train/images', data_val='./data/val/images', epochs=100, img_f='', lr=0.001, lrf=0.001, mask_f='')
结果保存在------> runs 目录
当前训练设备: cuda
数据加载使用线程数: 4
训练集个数:394	验证集个数:98
train:   0%|          | 0/99 [00:00<?, ?it/s]-----------train-----------[epoch:0/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.87it/s, dice=0.0348, iou=0.0177, loss=0.0842]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.50it/s, dice=0, iou=0, loss=0.0892]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00100000
train loss:0.1814 	 train mean iou:0.0183	 train mean dice:0.0359
val loss:0.0932 	 val mean iou:0.0000	 val mean dice:0.0000[epoch:1/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.94it/s, dice=0.287, iou=0.168, loss=0.0324]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.58it/s, dice=0.415, iou=0.262, loss=0.148]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00099975
train loss:0.0735 	 train mean iou:0.1044	 train mean dice:0.1803
val loss:0.0605 	 val mean iou:0.3080	 val mean dice:0.4684[epoch:2/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.94it/s, dice=0.457, iou=0.296, loss=0.0236]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.52it/s, dice=0.534, iou=0.364, loss=0.261]
learning rate:0.00099901
train loss:0.0594 	 train mean iou:0.3108	 train mean dice:0.4724
val loss:0.0621 	 val mean iou:0.3928	 val mean dice:0.5636[epoch:3/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.94it/s, dice=0.525, iou=0.356, loss=0.0339]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.58it/s, dice=0.614, iou=0.443, loss=0.153]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00099778
train loss:0.0534 	 train mean iou:0.3698	 train mean dice:0.5394
val loss:0.0445 	 val mean iou:0.4533	 val mean dice:0.6229[epoch:4/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.538, iou=0.368, loss=0.0418]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.58it/s, dice=0.61, iou=0.439, loss=0.47]
learning rate:0.00099606
train loss:0.0526 	 train mean iou:0.3817	 train mean dice:0.5502
val loss:0.0691 	 val mean iou:0.4654	 val mean dice:0.6347[epoch:5/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.572, iou=0.4, loss=0.187]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.44it/s, dice=0.588, iou=0.416, loss=0.376]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00099385
train loss:0.0496 	 train mean iou:0.4191	 train mean dice:0.5893
val loss:0.0599 	 val mean iou:0.4405	 val mean dice:0.6112[epoch:6/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.583, iou=0.411, loss=0.0304]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.57it/s, dice=0.667, iou=0.5, loss=0.0661]
learning rate:0.00099115
train loss:0.0473 	 train mean iou:0.4157	 train mean dice:0.5867
val loss:0.0483 	 val mean iou:0.5008	 val mean dice:0.6667[epoch:7/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.551, iou=0.381, loss=0.0357]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.65it/s, dice=0.687, iou=0.523, loss=0.32]
learning rate:0.00098797
train loss:0.0471 	 train mean iou:0.3781	 train mean dice:0.5477
val loss:0.0593 	 val mean iou:0.5040	 val mean dice:0.6696[epoch:8/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.619, iou=0.448, loss=0.0487]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.44it/s, dice=0.612, iou=0.441, loss=0.142]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00098431
train loss:0.0459 	 train mean iou:0.4625	 train mean dice:0.6311
val loss:0.0454 	 val mean iou:0.3574	 val mean dice:0.5226[epoch:9/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.6, iou=0.428, loss=0.0207]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.52it/s, dice=0.76, iou=0.613, loss=0.0626]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00098017
train loss:0.0449 	 train mean iou:0.4584	 train mean dice:0.6278
val loss:0.0428 	 val mean iou:0.6277	 val mean dice:0.7703[epoch:10/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.94it/s, dice=0.662, iou=0.495, loss=0.0304]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.43it/s, dice=0.674, iou=0.509, loss=0.186]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00097555
train loss:0.0432 	 train mean iou:0.5322	 train mean dice:0.6937
val loss:0.0495 	 val mean iou:0.5242	 val mean dice:0.6877[epoch:11/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.679, iou=0.514, loss=0.103]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.46it/s, dice=0.738, iou=0.585, loss=0.121]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00097047
train loss:0.0435 	 train mean iou:0.5127	 train mean dice:0.6773
val loss:0.0484 	 val mean iou:0.6131	 val mean dice:0.7595[epoch:12/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.676, iou=0.511, loss=0.0221]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.33it/s, dice=0.714, iou=0.555, loss=0.066]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00096492
train loss:0.0406 	 train mean iou:0.5087	 train mean dice:0.6738
val loss:0.0432 	 val mean iou:0.5400	 val mean dice:0.7011[epoch:13/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.696, iou=0.533, loss=0.0209]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.52it/s, dice=0.69, iou=0.527, loss=0.334]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00095892
train loss:0.0388 	 train mean iou:0.5277	 train mean dice:0.6898
val loss:0.0480 	 val mean iou:0.5596	 val mean dice:0.7170[epoch:14/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.609, iou=0.438, loss=0.227]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.53it/s, dice=0.71, iou=0.551, loss=0.216]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00095246
train loss:0.0427 	 train mean iou:0.4521	 train mean dice:0.6218
val loss:0.0408 	 val mean iou:0.4988	 val mean dice:0.6644[epoch:15/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.704, iou=0.543, loss=0.0964]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.45it/s, dice=0.508, iou=0.341, loss=0.171]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00094556
train loss:0.0415 	 train mean iou:0.5090	 train mean dice:0.6715
val loss:0.0455 	 val mean iou:0.3878	 val mean dice:0.5549[epoch:16/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.657, iou=0.489, loss=0.0191]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.50it/s, dice=0.627, iou=0.456, loss=0.39]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00093822
train loss:0.0426 	 train mean iou:0.4296	 train mean dice:0.5961
val loss:0.0530 	 val mean iou:0.4520	 val mean dice:0.6214[epoch:17/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.677, iou=0.512, loss=0.0802]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.33it/s, dice=0.728, iou=0.572, loss=0.162]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00093044
train loss:0.0386 	 train mean iou:0.4780	 train mean dice:0.6444
val loss:0.0504 	 val mean iou:0.5886	 val mean dice:0.7407[epoch:18/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.82it/s, dice=0.709, iou=0.549, loss=0.037]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.53it/s, dice=0.723, iou=0.566, loss=0.147]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00092224
train loss:0.0395 	 train mean iou:0.5334	 train mean dice:0.6948
val loss:0.0429 	 val mean iou:0.5431	 val mean dice:0.7034[epoch:19/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.709, iou=0.549, loss=0.0166]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.48it/s, dice=0.769, iou=0.625, loss=0.307]
learning rate:0.00091363
train loss:0.0380 	 train mean iou:0.5297	 train mean dice:0.6922
val loss:0.0468 	 val mean iou:0.6245	 val mean dice:0.7663[epoch:20/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.677, iou=0.512, loss=0.0246]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.54it/s, dice=0.754, iou=0.605, loss=0.161]
learning rate:0.00090460
train loss:0.0372 	 train mean iou:0.5491	 train mean dice:0.7075
val loss:0.0325 	 val mean iou:0.5803	 val mean dice:0.7338[epoch:21/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.734, iou=0.58, loss=0.03]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.48it/s, dice=0.767, iou=0.622, loss=0.222]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00089518
train loss:0.0359 	 train mean iou:0.5445	 train mean dice:0.7041
val loss:0.0331 	 val mean iou:0.6308	 val mean dice:0.7733[epoch:22/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.775, iou=0.633, loss=0.0185]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.53it/s, dice=0.795, iou=0.66, loss=0.0489]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00088537
train loss:0.0311 	 train mean iou:0.6291	 train mean dice:0.7721
val loss:0.0334 	 val mean iou:0.6434	 val mean dice:0.7813[epoch:23/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.735, iou=0.581, loss=0.0271]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.54it/s, dice=0.768, iou=0.623, loss=0.257]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00087518
train loss:0.0345 	 train mean iou:0.5402	 train mean dice:0.7004
val loss:0.0378 	 val mean iou:0.6493	 val mean dice:0.7872[epoch:24/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.751, iou=0.601, loss=0.0279]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.49it/s, dice=0.751, iou=0.601, loss=0.424]
learning rate:0.00086462
train loss:0.0317 	 train mean iou:0.6022	 train mean dice:0.7515
val loss:0.0477 	 val mean iou:0.6696	 val mean dice:0.8015[epoch:25/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.748, iou=0.598, loss=0.0312]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.36it/s, dice=0.74, iou=0.588, loss=0.237]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00085370
train loss:0.0352 	 train mean iou:0.5870	 train mean dice:0.7394
val loss:0.0416 	 val mean iou:0.6053	 val mean dice:0.7534[epoch:26/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.762, iou=0.615, loss=0.0105]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.32it/s, dice=0.764, iou=0.618, loss=0.233]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00084243
train loss:0.0350 	 train mean iou:0.6163	 train mean dice:0.7622
val loss:0.0378 	 val mean iou:0.6547	 val mean dice:0.7906[epoch:27/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.768, iou=0.624, loss=0.049]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.39it/s, dice=0.798, iou=0.664, loss=0.0638]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00083083
train loss:0.0306 	 train mean iou:0.6427	 train mean dice:0.7820
val loss:0.0301 	 val mean iou:0.6595	 val mean dice:0.7945[epoch:28/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.89it/s, dice=0.753, iou=0.604, loss=0.0897]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.42it/s, dice=0.79, iou=0.653, loss=0.418]
learning rate:0.00081889
train loss:0.0353 	 train mean iou:0.6500	 train mean dice:0.7871
val loss:0.0429 	 val mean iou:0.6929	 val mean dice:0.8184[epoch:29/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.775, iou=0.632, loss=0.0133]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.40it/s, dice=0.783, iou=0.644, loss=0.277]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00080665
train loss:0.0308 	 train mean iou:0.6426	 train mean dice:0.7821
val loss:0.0338 	 val mean iou:0.6590	 val mean dice:0.7940[epoch:30/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.724, iou=0.568, loss=0.0182]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.41it/s, dice=0.801, iou=0.667, loss=0.114]
learning rate:0.00079410
train loss:0.0347 	 train mean iou:0.6146	 train mean dice:0.7603
val loss:0.0301 	 val mean iou:0.6981	 val mean dice:0.8216[epoch:31/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.764, iou=0.618, loss=0.0476]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.41it/s, dice=0.846, iou=0.733, loss=0.246]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00078126
train loss:0.0269 	 train mean iou:0.6629	 train mean dice:0.7959
val loss:0.0357 	 val mean iou:0.7557	 val mean dice:0.8606[epoch:32/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.777, iou=0.635, loss=0.0253]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.40it/s, dice=0.799, iou=0.665, loss=0.242]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00076815
train loss:0.0317 	 train mean iou:0.6592	 train mean dice:0.7943
val loss:0.0305 	 val mean iou:0.6464	 val mean dice:0.7844[epoch:33/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.87it/s, dice=0.808, iou=0.678, loss=0.011]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.58it/s, dice=0.776, iou=0.634, loss=0.113]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00075477
train loss:0.0266 	 train mean iou:0.6881	 train mean dice:0.8152
val loss:0.0336 	 val mean iou:0.6222	 val mean dice:0.7670[epoch:34/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.89it/s, dice=0.819, iou=0.694, loss=0.0152]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.59it/s, dice=0.767, iou=0.623, loss=0.103]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00074114
train loss:0.0274 	 train mean iou:0.6748	 train mean dice:0.8054
val loss:0.0376 	 val mean iou:0.6323	 val mean dice:0.7744[epoch:35/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.88it/s, dice=0.83, iou=0.709, loss=0.0402]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.57it/s, dice=0.751, iou=0.602, loss=0.201]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00072727
train loss:0.0258 	 train mean iou:0.6924	 train mean dice:0.8172
val loss:0.0325 	 val mean iou:0.6234	 val mean dice:0.7678[epoch:36/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.8, iou=0.666, loss=0.0183]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.48it/s, dice=0.805, iou=0.674, loss=0.156]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00071318
train loss:0.0285 	 train mean iou:0.6383	 train mean dice:0.7787
val loss:0.0264 	 val mean iou:0.6985	 val mean dice:0.8223[epoch:37/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.8, iou=0.667, loss=0.0235]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.52it/s, dice=0.802, iou=0.67, loss=0.177]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00069888
train loss:0.0239 	 train mean iou:0.7212	 train mean dice:0.8378
val loss:0.0309 	 val mean iou:0.6214	 val mean dice:0.7657[epoch:38/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.89it/s, dice=0.836, iou=0.718, loss=0.00685]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.51it/s, dice=0.857, iou=0.75, loss=0.111]
learning rate:0.00068438
train loss:0.0225 	 train mean iou:0.7076	 train mean dice:0.8285
val loss:0.0256 	 val mean iou:0.7558	 val mean dice:0.8608[epoch:39/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.823, iou=0.699, loss=0.0215]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.54it/s, dice=0.852, iou=0.742, loss=0.103]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00066970
train loss:0.0231 	 train mean iou:0.7180	 train mean dice:0.8357
val loss:0.0238 	 val mean iou:0.7280	 val mean dice:0.8411[epoch:40/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.823, iou=0.699, loss=0.00844]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.50it/s, dice=0.848, iou=0.736, loss=0.0692]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00065485
train loss:0.0255 	 train mean iou:0.6789	 train mean dice:0.8084
val loss:0.0253 	 val mean iou:0.7462	 val mean dice:0.8544[epoch:41/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.855, iou=0.747, loss=0.00849]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.38it/s, dice=0.872, iou=0.774, loss=0.0719]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00063986
train loss:0.0202 	 train mean iou:0.7336	 train mean dice:0.8462
val loss:0.0182 	 val mean iou:0.7430	 val mean dice:0.8515[epoch:42/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.862, iou=0.758, loss=0.00707]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.48it/s, dice=0.858, iou=0.752, loss=0.0445]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00062472
train loss:0.0205 	 train mean iou:0.7617	 train mean dice:0.8646
val loss:0.0222 	 val mean iou:0.7416	 val mean dice:0.8507[epoch:43/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.88, iou=0.785, loss=0.0208]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.48it/s, dice=0.861, iou=0.756, loss=0.154]
learning rate:0.00060946
train loss:0.0169 	 train mean iou:0.7738	 train mean dice:0.8724
val loss:0.0236 	 val mean iou:0.7688	 val mean dice:0.8691[epoch:44/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.859, iou=0.753, loss=0.00897]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.52it/s, dice=0.886, iou=0.795, loss=0.145]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00059410
train loss:0.0225 	 train mean iou:0.7569	 train mean dice:0.8614
val loss:0.0206 	 val mean iou:0.7677	 val mean dice:0.8674[epoch:45/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.869, iou=0.768, loss=0.00463]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.51it/s, dice=0.844, iou=0.73, loss=0.0639]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00057864
train loss:0.0195 	 train mean iou:0.7563	 train mean dice:0.8611
val loss:0.0200 	 val mean iou:0.7664	 val mean dice:0.8676[epoch:46/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.85it/s, dice=0.87, iou=0.77, loss=0.0116]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.57it/s, dice=0.883, iou=0.791, loss=0.117]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00056310
train loss:0.0200 	 train mean iou:0.7826	 train mean dice:0.8779
val loss:0.0215 	 val mean iou:0.8068	 val mean dice:0.8921[epoch:47/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.881, iou=0.787, loss=0.00692]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.54it/s, dice=0.879, iou=0.784, loss=0.0711]
learning rate:0.00054751
train loss:0.0184 	 train mean iou:0.7881	 train mean dice:0.8814
val loss:0.0211 	 val mean iou:0.8116	 val mean dice:0.8959[epoch:48/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.882, iou=0.789, loss=0.0737]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.53it/s, dice=0.868, iou=0.767, loss=0.307]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00053186
train loss:0.0191 	 train mean iou:0.7954	 train mean dice:0.8859
val loss:0.0298 	 val mean iou:0.7899	 val mean dice:0.8825[epoch:49/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.89it/s, dice=0.874, iou=0.776, loss=0.0201]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.40it/s, dice=0.88, iou=0.786, loss=0.0492]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00051619
train loss:0.0187 	 train mean iou:0.7613	 train mean dice:0.8643
val loss:0.0225 	 val mean iou:0.7977	 val mean dice:0.8872[epoch:50/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.889, iou=0.801, loss=0.00498]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.52it/s, dice=0.876, iou=0.779, loss=0.0773]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00050050
train loss:0.0180 	 train mean iou:0.8037	 train mean dice:0.8911
val loss:0.0208 	 val mean iou:0.8059	 val mean dice:0.8925[epoch:51/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.89it/s, dice=0.883, iou=0.791, loss=0.00468]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.54it/s, dice=0.888, iou=0.798, loss=0.0844]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00048481
train loss:0.0188 	 train mean iou:0.7785	 train mean dice:0.8752
val loss:0.0173 	 val mean iou:0.7890	 val mean dice:0.8820[epoch:52/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.89it/s, dice=0.889, iou=0.801, loss=0.02]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.55it/s, dice=0.868, iou=0.767, loss=0.0538]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00046914
train loss:0.0167 	 train mean iou:0.8123	 train mean dice:0.8963
val loss:0.0172 	 val mean iou:0.7676	 val mean dice:0.8681[epoch:53/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.898, iou=0.816, loss=0.104]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.49it/s, dice=0.873, iou=0.775, loss=0.0426]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00045349
train loss:0.0168 	 train mean iou:0.8257	 train mean dice:0.9045
val loss:0.0214 	 val mean iou:0.8032	 val mean dice:0.8908[epoch:54/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.869, iou=0.769, loss=0.00437]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.54it/s, dice=0.883, iou=0.79, loss=0.048]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00043790
train loss:0.0201 	 train mean iou:0.7538	 train mean dice:0.8596
val loss:0.0182 	 val mean iou:0.7921	 val mean dice:0.8839[epoch:55/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.89, iou=0.802, loss=0.0137]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.53it/s, dice=0.898, iou=0.815, loss=0.114]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00042236
train loss:0.0185 	 train mean iou:0.8078	 train mean dice:0.8936
val loss:0.0188 	 val mean iou:0.8498	 val mean dice:0.9187[epoch:56/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.89it/s, dice=0.883, iou=0.79, loss=0.0239]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.50it/s, dice=0.896, iou=0.811, loss=0.0694]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00040690
train loss:0.0188 	 train mean iou:0.7808	 train mean dice:0.8768
val loss:0.0171 	 val mean iou:0.8254	 val mean dice:0.9043[epoch:57/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.891, iou=0.803, loss=0.00324]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.45it/s, dice=0.888, iou=0.798, loss=0.0399]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00039154
train loss:0.0173 	 train mean iou:0.8041	 train mean dice:0.8914
val loss:0.0149 	 val mean iou:0.8090	 val mean dice:0.8944[epoch:58/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.894, iou=0.809, loss=0.0032]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.61it/s, dice=0.887, iou=0.797, loss=0.06]
learning rate:0.00037628
train loss:0.0154 	 train mean iou:0.7995	 train mean dice:0.8878
val loss:0.0166 	 val mean iou:0.8245	 val mean dice:0.9037[epoch:59/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.905, iou=0.827, loss=0.00593]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.47it/s, dice=0.908, iou=0.832, loss=0.0438]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00036114
train loss:0.0144 	 train mean iou:0.8198	 train mean dice:0.9009
val loss:0.0139 	 val mean iou:0.8406	 val mean dice:0.9133[epoch:60/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.899, iou=0.816, loss=0.186]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.42it/s, dice=0.892, iou=0.805, loss=0.0514]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00034615
train loss:0.0163 	 train mean iou:0.8271	 train mean dice:0.9053
val loss:0.0160 	 val mean iou:0.8337	 val mean dice:0.9091[epoch:61/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.88it/s, dice=0.888, iou=0.798, loss=0.022]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.49it/s, dice=0.901, iou=0.819, loss=0.0437]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00033130
train loss:0.0165 	 train mean iou:0.8006	 train mean dice:0.8892
val loss:0.0157 	 val mean iou:0.8006	 val mean dice:0.8884[epoch:62/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.894, iou=0.808, loss=0.0592]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.47it/s, dice=0.881, iou=0.788, loss=0.0476]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00031662
train loss:0.0165 	 train mean iou:0.7903	 train mean dice:0.8827
val loss:0.0202 	 val mean iou:0.7907	 val mean dice:0.8829[epoch:63/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.898, iou=0.815, loss=0.00335]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.61it/s, dice=0.902, iou=0.821, loss=0.0388]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00030212
train loss:0.0141 	 train mean iou:0.8146	 train mean dice:0.8978
val loss:0.0155 	 val mean iou:0.8363	 val mean dice:0.9105[epoch:64/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.904, iou=0.825, loss=0.00387]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.62it/s, dice=0.894, iou=0.808, loss=0.0617]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00028782
train loss:0.0142 	 train mean iou:0.8385	 train mean dice:0.9121
val loss:0.0168 	 val mean iou:0.8334	 val mean dice:0.9091[epoch:65/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.9, iou=0.818, loss=0.00276]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.49it/s, dice=0.906, iou=0.829, loss=0.0362]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00027373
train loss:0.0152 	 train mean iou:0.8226	 train mean dice:0.9027
val loss:0.0154 	 val mean iou:0.8413	 val mean dice:0.9134[epoch:66/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.91, iou=0.834, loss=0.0309]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.44it/s, dice=0.911, iou=0.837, loss=0.0362]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00025986
train loss:0.0138 	 train mean iou:0.8325	 train mean dice:0.9084
val loss:0.0156 	 val mean iou:0.8443	 val mean dice:0.9155[epoch:67/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.918, iou=0.848, loss=0.00699]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.49it/s, dice=0.911, iou=0.837, loss=0.0526]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00024623
train loss:0.0121 	 train mean iou:0.8453	 train mean dice:0.9161
val loss:0.0150 	 val mean iou:0.8549	 val mean dice:0.9217[epoch:68/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.905, iou=0.826, loss=0.00609]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.51it/s, dice=0.888, iou=0.798, loss=0.0435]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00023285
train loss:0.0135 	 train mean iou:0.8385	 train mean dice:0.9121
val loss:0.0191 	 val mean iou:0.8243	 val mean dice:0.9033[epoch:69/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.899, iou=0.816, loss=0.0075]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.46it/s, dice=0.901, iou=0.819, loss=0.0631]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00021974
train loss:0.0168 	 train mean iou:0.8066	 train mean dice:0.8929
val loss:0.0174 	 val mean iou:0.8223	 val mean dice:0.9025[epoch:70/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.91, iou=0.835, loss=0.00366]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.49it/s, dice=0.885, iou=0.794, loss=0.031]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00020690
train loss:0.0128 	 train mean iou:0.8380	 train mean dice:0.9118
val loss:0.0175 	 val mean iou:0.8103	 val mean dice:0.8950[epoch:71/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.896, iou=0.811, loss=0.0154]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.50it/s, dice=0.912, iou=0.839, loss=0.0637]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00019435
train loss:0.0156 	 train mean iou:0.7958	 train mean dice:0.8862
val loss:0.0134 	 val mean iou:0.8227	 val mean dice:0.9024[epoch:72/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.911, iou=0.836, loss=0.0134]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.44it/s, dice=0.902, iou=0.822, loss=0.0514]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00018211
train loss:0.0146 	 train mean iou:0.8412	 train mean dice:0.9137
val loss:0.0169 	 val mean iou:0.8304	 val mean dice:0.9072[epoch:73/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.908, iou=0.831, loss=0.00615]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.63it/s, dice=0.908, iou=0.832, loss=0.0415]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00017017
train loss:0.0144 	 train mean iou:0.8351	 train mean dice:0.9101
val loss:0.0127 	 val mean iou:0.8422	 val mean dice:0.9143[epoch:74/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.913, iou=0.841, loss=0.0146]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.47it/s, dice=0.908, iou=0.832, loss=0.0612]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00015857
train loss:0.0121 	 train mean iou:0.8368	 train mean dice:0.9112
val loss:0.0136 	 val mean iou:0.8430	 val mean dice:0.9148[epoch:75/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.921, iou=0.854, loss=0.00527]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.45it/s, dice=0.902, iou=0.821, loss=0.0521]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00014730
train loss:0.0123 	 train mean iou:0.8527	 train mean dice:0.9203
val loss:0.0170 	 val mean iou:0.8436	 val mean dice:0.9151[epoch:76/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.92, iou=0.851, loss=0.022]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.47it/s, dice=0.924, iou=0.859, loss=0.0556]
learning rate:0.00013638
train loss:0.0121 	 train mean iou:0.8530	 train mean dice:0.9207
val loss:0.0112 	 val mean iou:0.8730	 val mean dice:0.9321[epoch:77/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.91, iou=0.835, loss=0.00476]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.49it/s, dice=0.91, iou=0.835, loss=0.0338]
learning rate:0.00012582
train loss:0.0153 	 train mean iou:0.8392	 train mean dice:0.9125
val loss:0.0126 	 val mean iou:0.8553	 val mean dice:0.9217[epoch:78/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.909, iou=0.834, loss=0.00865]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.53it/s, dice=0.918, iou=0.848, loss=0.0374]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00011563
train loss:0.0134 	 train mean iou:0.8401	 train mean dice:0.9131
val loss:0.0150 	 val mean iou:0.8528	 val mean dice:0.9205[epoch:79/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.919, iou=0.85, loss=0.00785]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.45it/s, dice=0.923, iou=0.857, loss=0.0325]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00010582
train loss:0.0124 	 train mean iou:0.8532	 train mean dice:0.9207
val loss:0.0123 	 val mean iou:0.8643	 val mean dice:0.9272[epoch:80/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.916, iou=0.846, loss=0.00668]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.51it/s, dice=0.917, iou=0.846, loss=0.034]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00009640
train loss:0.0121 	 train mean iou:0.8513	 train mean dice:0.9196
val loss:0.0120 	 val mean iou:0.8515	 val mean dice:0.9197[epoch:81/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.914, iou=0.842, loss=0.00393]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.46it/s, dice=0.914, iou=0.841, loss=0.0417]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00008737
train loss:0.0134 	 train mean iou:0.8594	 train mean dice:0.9243
val loss:0.0132 	 val mean iou:0.8529	 val mean dice:0.9206[epoch:82/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.914, iou=0.842, loss=0.011]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.49it/s, dice=0.911, iou=0.836, loss=0.0316]
learning rate:0.00007876
train loss:0.0117 	 train mean iou:0.8409	 train mean dice:0.9135
val loss:0.0156 	 val mean iou:0.8319	 val mean dice:0.9082[epoch:83/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.924, iou=0.859, loss=0.00489]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.45it/s, dice=0.923, iou=0.858, loss=0.0323]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00007056
train loss:0.0112 	 train mean iou:0.8487	 train mean dice:0.9181
val loss:0.0141 	 val mean iou:0.8739	 val mean dice:0.9327[epoch:84/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.92, iou=0.852, loss=0.00831]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.50it/s, dice=0.897, iou=0.813, loss=0.0397]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00006278
train loss:0.0115 	 train mean iou:0.8599	 train mean dice:0.9246
val loss:0.0145 	 val mean iou:0.7953	 val mean dice:0.8844[epoch:85/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.921, iou=0.853, loss=0.012]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.48it/s, dice=0.918, iou=0.848, loss=0.0379]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00005544
train loss:0.0113 	 train mean iou:0.8437	 train mean dice:0.9151
val loss:0.0145 	 val mean iou:0.8553	 val mean dice:0.9220[epoch:86/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.924, iou=0.859, loss=0.0162]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.46it/s, dice=0.917, iou=0.846, loss=0.0319]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00004854
train loss:0.0105 	 train mean iou:0.8651	 train mean dice:0.9276
val loss:0.0120 	 val mean iou:0.8221	 val mean dice:0.9022[epoch:87/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.918, iou=0.848, loss=0.00446]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.51it/s, dice=0.918, iou=0.848, loss=0.0463]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00004208
train loss:0.0119 	 train mean iou:0.8557	 train mean dice:0.9222
val loss:0.0137 	 val mean iou:0.8377	 val mean dice:0.9113[epoch:88/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.92it/s, dice=0.916, iou=0.845, loss=0.0165]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.63it/s, dice=0.919, iou=0.851, loss=0.0361]
learning rate:0.00003608
train loss:0.0118 	 train mean iou:0.8311	 train mean dice:0.9076
val loss:0.0123 	 val mean iou:0.8688	 val mean dice:0.9297[epoch:89/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.916, iou=0.844, loss=0.0188]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.44it/s, dice=0.908, iou=0.832, loss=0.0455]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00003053
train loss:0.0122 	 train mean iou:0.8360	 train mean dice:0.9105
val loss:0.0137 	 val mean iou:0.8420	 val mean dice:0.9142[epoch:90/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.90it/s, dice=0.923, iou=0.857, loss=0.0135]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.47it/s, dice=0.928, iou=0.866, loss=0.0499]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00002545
train loss:0.0111 	 train mean iou:0.8468	 train mean dice:0.9168
val loss:0.0117 	 val mean iou:0.8696	 val mean dice:0.9301[epoch:91/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.93it/s, dice=0.925, iou=0.86, loss=0.0101]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.48it/s, dice=0.924, iou=0.858, loss=0.0609]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00002083
train loss:0.0117 	 train mean iou:0.8582	 train mean dice:0.9237
val loss:0.0142 	 val mean iou:0.8752	 val mean dice:0.9334[epoch:92/99]
train: 100%|██████████| 99/99 [00:26<00:00,  3.81it/s, dice=0.924, iou=0.859, loss=0.0144]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.43it/s, dice=0.908, iou=0.832, loss=0.0489]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00001669
train loss:0.0110 	 train mean iou:0.8616	 train mean dice:0.9256
val loss:0.0134 	 val mean iou:0.8631	 val mean dice:0.9264[epoch:93/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.92, iou=0.852, loss=0.00768]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.47it/s, dice=0.919, iou=0.85, loss=0.0536]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00001303
train loss:0.0112 	 train mean iou:0.8504	 train mean dice:0.9191
val loss:0.0111 	 val mean iou:0.8171	 val mean dice:0.8986[epoch:94/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.924, iou=0.859, loss=0.00417]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.55it/s, dice=0.917, iou=0.846, loss=0.0214]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00000985
train loss:0.0112 	 train mean iou:0.8596	 train mean dice:0.9245
val loss:0.0126 	 val mean iou:0.8453	 val mean dice:0.9161[epoch:95/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.919, iou=0.851, loss=0.0148]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.59it/s, dice=0.921, iou=0.854, loss=0.0324]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00000715
train loss:0.0111 	 train mean iou:0.8576	 train mean dice:0.9233
val loss:0.0108 	 val mean iou:0.8619	 val mean dice:0.9258[epoch:96/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.89it/s, dice=0.917, iou=0.846, loss=0.0039]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.55it/s, dice=0.923, iou=0.857, loss=0.0484]
learning rate:0.00000494
train loss:0.0138 	 train mean iou:0.8633	 train mean dice:0.9266
val loss:0.0134 	 val mean iou:0.8777	 val mean dice:0.9348[epoch:97/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.88it/s, dice=0.922, iou=0.855, loss=0.0133]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.30it/s, dice=0.904, iou=0.825, loss=0.0454]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00000322
train loss:0.0115 	 train mean iou:0.8417	 train mean dice:0.9140
val loss:0.0165 	 val mean iou:0.8324	 val mean dice:0.9085[epoch:98/99]
train: 100%|██████████| 99/99 [00:26<00:00,  3.70it/s, dice=0.922, iou=0.855, loss=0.00323]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.52it/s, dice=0.915, iou=0.843, loss=0.0304]
train:   0%|          | 0/99 [00:00<?, ?it/s]learning rate:0.00000199
train loss:0.0111 	 train mean iou:0.8663	 train mean dice:0.9283
val loss:0.0123 	 val mean iou:0.8549	 val mean dice:0.9217[epoch:99/99]
train: 100%|██████████| 99/99 [00:25<00:00,  3.91it/s, dice=0.925, iou=0.861, loss=0.0134]
valid: 100%|██████████| 25/25 [00:05<00:00,  4.42it/s, dice=0.931, iou=0.87, loss=0.0378]
learning rate:0.00000125
train loss:0.0105 	 train mean iou:0.8607	 train mean dice:0.9251
val loss:0.0105 	 val mean iou:0.8805	 val mean dice:0.9358train finish!
验证集上表现最好的epoch为: 99进程已结束，退出代码为 0

8.代码

如下：基于Unet融合SAM模型point提示推理医学图像分割项目：腹部多脏器分割资源-CSDN文库

其他改进：AI 改进系列_听风吹等浪起的博客-CSDN博客

改进系列_听风吹等浪起的博客-CSDN博客

http://www.xdnf.cn/news/9261.html

相关文章：

  • 【北京盈达科技】GEO优化：引领AI时代内容霸权，重塑行业生态
  • 思澈科技助力Keep Watch Pilot 1：重新定义智能运动手表体验
  • React 虚拟dom
  • ROS2 robot控制学习(一)
  • 自然语言×数据集成新范式：SeaTunnel MCP深度解读 | 附视频讲解
  • 重新安装解决mac vscode点击不能跳转问题
  • 树莓派（Raspberry Pi）安装Docker教程
  • LabVIEW软件开发过程中如何保证软件的质量？
  • 大数据-272 Spark MLib - 基础介绍 机器学习算法 线性回归
  • openresty如何禁止海外ip访问
  • 【git】git rebase 和 git pull区别？
  • NSSCTF [NISACTF 2022]ezheap
  • 微信小程序的软件测试用例编写指南及示例--性能测试用例
  • 使用Gemini, LangChain, Gradio打造一个书籍推荐系统 （第三部分）
  • 查服务器信息 常用的一些命令 =^^ =
  • 共现矩阵的SVD降维与低维词向量计算详解
  • AI 智能体的那些事—架构设计关键点
  • 【Java实战】集合排序方法与长度获取方法辨析（易懂版）
  • 11.Java I/O 流：文件读写与数据持久化​
  • 夏季用电高峰如何防患于未“燃”？电力测温技术守护城市生命线
  • 使用 Redis 作为向量数据库
  • 5G 核心网 UE 状态深度剖析：机制、迁移与演进
  • 新版Chrome浏览器加载eDrawings 3D Viewer控件网页查看DWG、DXF
  • 利用Tushare+pyEcharts进行沪深证券数据采集与分析
  • 单向循环链表与双向链表
  • 洗鞋店干洗店线上预约管理系统；
  • 【OS安装与使用】part7-ubuntu22.04LTS 的 docker 安装与使用(实例：MTransServer服务部署)
  • AI辅助写作 从提笔难到高效创作的智能升级
  • WPF事件处理器+x名称空间
  • 基于多流特征融合与领域知识整合的CNN-xLSTM-xAtt网络用于光电容积脉搏波信号的无创血压估计【代码已复现】