SegEv: semantic segmentation performance verification and evaluation software

With the widespread application of semantic segmentation technology in fields such as remote sensing and industrial inspection, the evaluation of model performance and visualization of training processes have become key issues. This paper develops an integrated evaluation software based on PyQt5 and TensorBoard, which supports the calculation of eight metrics including Precision, Recall, F1, Accuracy, mPA, mIoU, Dice, ROC, and PR and provides functions such as multi-algorithm comparison and batch processing. Through TensorBoard, the software enables the visualization of model architectures, feature maps, heatmaps, and loss maps, intuitively displaying the differences between segmentation results and ground truth labels to assist in parameter optimization. With its modular design, the software combines both evaluation and visualization capabilities, providing efficient tool support for the development and deployment of segmentation models.