Curriculum Learning, as a deep learning training strategy, has proven its effectiveness across many Natural Language Processing (NLP) and Multi-modal tasks. However, traditional Curriculum Learning is often instance-level, which requires manually classifying each instance in the data into different difficulty level. This progress is not only time-consuming but also costly in labor. To outcome this challenge, we introduce an innovative training approach named Sequential Augmentation with Curriculum Learning (SACL). SACL is a curriculum learning approach in dataset-level, which utilizes various data augmentation techniques to generate data in various difficulty levels and feeds this data to the model in an order of increasing difficulty. This approach effectively avoids the process of manually classifying data in traditional curriculum learning which greatly reduces the consumption of time and resources. To validate our method, we conduct experiments on the MET-MEME dataset for a multimodal emotion classification task and deeply analyze the impact of curriculum learning on the training results. The results prove the effectiveness of our method.