— Semantic segmentation remains a fundamental challenge in computer vision, where the choice and weighting of loss functions significantly impact model performance. This study presents a comprehensive comparative analysis of individual versus combined loss functions with systematic weight ablation for semantic segmentation using modified Attention U-Net and DeepLabV3+ architectures on the Cityscapes dataset. We systematically evaluate seven weight configurations across three loss components (Cross-Entropy, Dice, Focal) through rigorous ablation studies, and validate our findings across two architectures to ensure generalizability. Through extensive experimentation across 20 epochs with 2,975 training and 500 validation images, our results demonstrate that the Dice-dominant weighting configuration (0.5:1.0:0.5 for CE:Dice:Focal) achieves superior performance with 57.83% mean Intersection over Union (mIoU) on Attention U-Net and 58.35% mIoU on DeepLabV3+, representing 7.78% improvement over the best individual loss function. Comprehensive ablation studies reveal that weight configuration critically affects performance, with Dice-dominant weighting consistently outperforming equal weighting (55.59% mIoU) and individual loss functions. Qualitative analysis demonstrates substantial improvements in boundary delineation and small object detection, with boundary IoU improving by 1.41% and challenging class performance (trucks, pedestrians) improving by 5 – 21%. Statistical analysis reveals that Cross-Entropy provides the most efficient training with a 75.4% loss reduction, while Dice loss exhibits convergence challenges, resulting in only a 34.5% reduction. Our findings conclusively demonstrate that optimized combined loss function weighting achieves better segmentation performance than both individual approaches and naive equal weighting strategies, with consistent improvements across different network architectures.
W. Swastika· Journal of Image and Graphic...· 0 citations
Background: Pneumonia remains a leading cause of mortality worldwide, with chest X-ray serving as the primary diagnostic tool. However, manual interpretation is subject to inter-observer variability, and existing deep learning models often require substantial computational resources that limit deployment in resource-constrained clinical environments.
Objective: This study aimed to develop Light CNN, a novel lightweight convolutional neural network that integrates depthwise separable convolution, inverted residual blocks, channel shuffle mechanism, and lightweight attention for efficient and accurate pneumonia classification from chest X-ray images.
Methods: Light CNN was designed with seven progressive feature extraction stages that incorporate the four aforementioned optimization techniques. The model was trained and evaluated on the publicly available Chest X-Ray Images (Pneumonia) dataset from Kaggle, comprising 5,856 images stratified into training (70%), validation (15%), and test (15%) subsets with patient-level splitting to prevent data leakage.
Preprocessing included CLAHE contrast enhancement, normalization, and data augmentation. Training employed the AdamW optimizer with cosine annealing scheduling and class-weighted cross-entropy loss over 50 epochs. The performance of Light CNN was benchmarked against three baseline models — MobileNetV2 (2.23 M parameters), ResNet-18 (11.18 M parameters), and EfficientNet-B0 (4.01 M parameters) — using identical preprocessing and training protocols. Evaluation metrics included accuracy, precision, recall, F1-score, AUC-ROC, parameter count, model size, and inference time.
Results: LightCNN achieved 95.56% accuracy, 0.9556 recall, 0.9584 precision, 0.9562 F1-score, and 0.9875 AUC-ROC on the test set, outperforming all baseline models. The model contains 2.52 million parameters (9.63 MB), representing a 77.4% reduction compared to ResNet-18, with an inference time of 0.25 ms per image — approximately four times faster than the nearest competitor. Ablation study results confirmed that each architectural component contributed incrementally to overall performance; depth wise separable convolution provided the largest efficiency gain, and inverted residual blocks contributed the most substantial accuracy improvement.
Conclusion: Light CNN demonstrates that systematic integration of lightweight architectural techniques can achieve clinically relevant diagnostic performance with minimal computational overhead, supporting its potential deployment in mobile and edge computing scenarios for point-of-care pneumonia diagnosis.
W. Swastika, Heri Kristianto, Paulus Lucky Tirma Irawan et al.· Health Sciences Investigatio...· 0 citations
