Image Captioning with CLIP and GPT-2 — Multimodal deep learning model integrating CLIP vision encoder and GPT-2 text decoder for image-to-text generation. Trained on 30K+ Flickr images, achieving BLEU-4 of 5.28% and CIDEr of 37.76%, outperforming CNN+LSTM baseline by 12%.