Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore cutting-edge research in computer vision through this advanced graduate-level course that examines the intersection of visual and language models, diffusion models, and AI safety. Delve into foundational concepts starting with CLIP and visual-language model architectures, progressing through comprehensive coverage of diffusion models for image and video generation. Investigate critical safety and bias issues in modern AI systems through analysis of recent research papers addressing object hallucinations in large vision-language models, bias detection in text-to-image generators, and safety evaluation frameworks for multimodal systems. Study methods for mitigating harmful content generation, including techniques for removing NSFW concepts from vision-language models and training-free safety guards for generative systems. Examine privacy concerns through research on private attribute inference from images and explore counterfactual approaches to addressing intersectional social bias. Learn about advanced decoding techniques for reducing hallucinations and discover approaches for selectively forgetting problematic concepts in diffusion models. Gain hands-on experience with state-of-the-art safety benchmarks and evaluation suites designed for multimodal large language models, preparing you to address the complex challenges of deploying vision-AI systems responsibly in real-world applications.
