Fine-Tuned Models, main dataset, and external dataset for brain tumor classification using SVM+HOG, ResNet18, ViT-B/16, and SimCLR

1: paper title : Trade-off Analysis of Classical and Deep Learning Models for Robust Brain Tumor Detection. 2: Please see the GitHub repository for code and usage instructions: https://github.com/yangzi334/BrainTumorClassification 3: uploading contents include: 3.1 please review all training and evaluatoin models associated with main brain tumor dataset (listed below): base_Vit_B_16_new.zip best_vit_overall.pth — Final fine-tuned ViT-B/16 model weights with best overall validation performance. best_vit_run1.pth — Best checkpoint for ViT-B/16 from run 1 (different random seed or split). best_vit_run2.pth — Best checkpoint for ViT-B/16 from run 2. best_vit_run3.pth — Best checkpoint for ViT-B/16 from run 3. base_resnet18_new.zip best_resnet18_overall.pth — Final fine-tuned resnet18 model weights with best overall validation performance. best_resnet18_run1.pth — Best checkpoint for resnet18 from run 1 (different random seed or split). best_resnet18_run2.pth — Best checkpoint for resnet18 from run 2. best_resnet18_run3.pth — Best checkpoint for resnet18 from run 3. base_simclr_new.zip best_simclr_overall.pth — Final fine-tuned simclr model weights with best overall validation performance. best_simclr_run1.pth — Best checkpoint for simclr from run 1 (different random seed or split). best_simclr_run2.pth — Best checkpoint for simclr from run 2. best_simclr_run3.pth — Best checkpoint for simclr from run 3. best_simclr_encoder_only.pth — SimCLR encoder weights only (trained with contrastive learning, no classifier head). Can be used for downstream tasks base_SVN_HOG_new.zip best_svm_hog_model.pkl — Final support vector machine (SVM) model trained with HOG features. Saved using joblib for easy loading and reuse. base tumor data.zip Contains the full brain tumor image dataset split into training, validation, and test sets: Each folder (train/, valid/, test/) includes: A COCO-format annotation file named _annotations.coco.json In the train folder, it includes fixed_annotations.coco.json Corresponding brain MRI image files in .jpg format 3.2 please review all training and evaluatoin models associated with external brain tumor dataset (listed below): base_Vit_B_16_new_extension.zip best_vit_extension_overall.pth — Final fine-tuned ViT-B/16 model weights with best overall validation performance. best_vit_extension_run1.pth — Best checkpoint for ViT-B/16 from run 1 (different random seed or split). best_vit_extension_run2.pth — Best checkpoint for ViT-B/16 from run 2. best_vit_extension_run3.pth — Best checkpoint for ViT-B/16 from run 3. base_resnet18_new_extension.zip best_resnet18_extension_overall.pth — Final fine-tuned resnet18 model weights with best overall validation performance. best_resnet18_extension_run1.pth — Best checkpoint for resnet18 from run 1 (different random seed or split). best_resnet18_extension_run2.pth — Best checkpoint for resnet18 from run 2. best_resnet18_extension_run3.pth — Best checkpoint for resnet18 from run 3. base_simclr_new_extension.zip best_simclr_extension_overall.pth — Final fine-tuned simclr model weights with best overall validation performance. best_simclr_extension_run1.pth — Best checkpoint for simclr from run 1 (different random seed or split). best_simclr_extension_run2.pth — Best checkpoint for simclr from run 2. best_simclr_extension_run3.pth — Best checkpoint for simclr from run 3. best_simclr_encoder_only.pth — SimCLR encoder weights only (trained with contrastive learning, no classifier head). Can be used for downstream tasks base_SVN_HOG_new_extension.zip best_svm_hog_model_extension.pkl — Final support vector machine (SVM) model trained with HOG features. Saved using joblib for easy loading and reuse. base tumor data extension source data.zip It's the external source dataset from the open-source on Figshare. This independent dataset comprises 2872 images with four subsets, including 828 glioma tumor images, 822 meningioma tumor images, 827 pituitary tumor images, and 395 no tumor images. base tumor data extension after preprocessing.zip It's the external dataset after completing the data preprocessing, we combined three tumor types - glioma, meiningioma, and pituitary into a single tumor class, then split the 2477 tumor images and 395 no-tumor images into train, validation, and test sets using the ratio of 0.7, 0.15, 0.15, saving them into the corresponding folders. It converted the 'base tumor data extension source' to the following structure: ‘..\train\tumor’ ‘..\validation\tumor’ ‘..\test\tumor’ ‘..\train\no-tumor’ ‘..\validation\no-tumor’ ‘..\test\no-tumor’