Apple slice grading is useful in post-harvest operations for sorting, grading, packaging, labeling, processing, storage, transportation, and meeting market demand and consumer preferences. Proper grading of apple slices can help ensure the quality, safety, and marketability of the final products, contributing to the post-harvest operations of the overall success of the apple industry. The article aims to create a convolutional neural network (CNN) model to classify images of apple slices after immersing them in atmospheric plasma at two different pressures (1 and 5 atm) and two different immersion times (3 and again 6 min) once and in filtered water based on the hardness of the slices using the k-Nearest Neighbors (KNN), Tree, Support Vector Machine (SVM), and Artificial Neural Network (ANN) algorithms. The results showed an inverse relationship between the storage period and the hardness of the apple slices, with the average hardness values gradually decreasing from 4.33 (day 1) to 3.37 (day 5). Treatment with atmospheric plasma at a pressure of 5 atm and an immersion time of 3 min gave the best results for maintaining the hardness of the slices during the storage period, recording values of 4.85 (first day) and 3.68 (fifth day), outperforming other treatments. The average improvement rate was 23.09% over five consecutive days. Regarding the CNN algorithms, the ANN algorithm achieved the highest classification accuracy of 97%, while the Tree algorithm achieved the lowest accuracy of 88.7%. The KNN and SVM algorithms achieved classification accuracies of 94.7% and 95.1%, respectively. The study demonstrated the possibility of using a CNN to classify apple slices based on the degree of hardness. Furthermore, the application of atmospheric plasma at 5 atmospheres with a 3-min immersion improves the firmness of the apple slices by inhibiting degradative enzymes while preserving the cellular structure and tissue quality.