Hey mate , one of the doubts i have with regards to building classification models with VLMs is the fact that wouldnt finetuning resnets or efficientnets (basically a CNN based model ) enough for these types of tasks.
Firstly need to understand what VLM ( Vision-Language Models) are. It's combining both Computer Vision and Natural Language Processing (NLP) to generate an output to describe photos or answer questions about visuals. While input can be Image, video, or text. So basically, they are using the Computer Vision functionality to an advanced level. So, Computer Vision won't be enough for these tasks.
Hey mate , one of the doubts i have with regards to building classification models with VLMs is the fact that wouldnt finetuning resnets or efficientnets (basically a CNN based model ) enough for these types of tasks.
What advantage does a VLM provide ?
VLM is not a replacement for CNN. They solve different types of problems.
Can you elaborate more?
Firstly need to understand what VLM ( Vision-Language Models) are. It's combining both Computer Vision and Natural Language Processing (NLP) to generate an output to describe photos or answer questions about visuals. While input can be Image, video, or text. So basically, they are using the Computer Vision functionality to an advanced level. So, Computer Vision won't be enough for these tasks.