The proper classification of harmful brain activity, particularly seizures, is critical in the diagnosis and treatment of seizures and other neurological disorders. EEG signals are critical indicators of the brain activity. This work offers a thorough method for categorizing seizures, lateralized rhythmic delta activity (LRDA), generalized rhythmic delta activity (GRDA), generalized periodic discharges (GPDs), and lateralized periodic discharges (LPDs) through the analysis of electroencephalograms (EEGs) data. We deploy here machine learning model inception-v3 that makes use of convolutional neural networks (CNNs) specifically, customized for two-dimensional time-series data. Our approach consists of preprocessing steps to improve signal quality, feature extraction to identify the characteristic indicators of seizure activity, and a strong classification algorithm that achieves high accuracy in differentiating between normal and abnormal brain activity. With 94.91% accuracy, our findings show a notable increase in the rates of seizure detection, opening the door for early intervention and individualized patient treatment which eventually help the patients to recover from their illness.