In this workshop, we cover the usage of Machine Learning for Hackathons. Specifically, we focus on pre-trained ML models, as they are the most practical and time-efficient solution for the fast pace of Hackathons.
You can find the workshop slide deck here.
For projects that involve text data, using pre-trained models like BERT, GPT, or spaCy can save time and provide strong baselines for tasks such as text classification, sentiment analysis, and entity recognition. Tools like Hugging Face's Transformers library offer easy access to these models and provide simple APIs to quickly integrate them into your projects.
You can find relevant materials in the nlp/ folder.
For projects that involve image or video data, pre-trained models like ResNet, VGG, or YOLO can be incredibly effective for tasks like image classification, object detection, and segmentation. Platforms such as TensorFlow Hub and PyTorch Hub provide ready-to-use models that can be fine-tuned with your dataset, allowing you to leverage state-of-the-art performance with minimal setup.
You can find relevant materials in the cv/ folder.