Online learning might be difficult for students with Attention-Deficit/Hyperactivity Disorder (ADHD). But most of instructional videos aren't produced with their needs, which might reduce learning results. To solve this, the ADHD-Aware Classification Approach (AAC), a deep learning-based approach created to evaluate the suitability of online educational video frames for students with (ADHD). The three phases of the suggested approach are preprocessing, training, and prediction with explainability. From the ADHD Online Instructor (AOI) dataset, a balanced dataset of 3,000 images, 1,500 suitable and 1,500 unsuitable was taken out and augmented using multiple techniques to enhance generalization. A MobileNetV2 architecture, enhanced with a Squeeze-and-Excitation (SE) block, was employed for classification. The model showed stability with a 98.3% test accuracy, strong precision, recall, and F1-score. Grad-CAM visualizations verified that the model's focus matched important teaching elements like the movement of the instructor's hands. The suggested model's better efficiency and ability for generalization were shown by comparison with ResNet50, VGG16, and InceptionV3. For students with ADHD, this activity offers a dependable and understandable way to improve information accessibility.