Identifying Wrongly Predicted Samples: A Method for Active Learning