YOLO-World VS CLIP

2024 is shaping up to be the year of multimodal machine learning. From real-time text-to-image models and open-world vocabulary models to multimodal large language models like GPT-4V and Gemini Pro Vision, AI is primed for an unprecedented array of interactive multimodal applications and experiences.

Post author here. I like this idea. I plan to explore it and make a more generic solution. I'd love to have a point-and-click interface for annotating scenes.

For example, I'd like to be able to click on pieces of coffee equipment in a photo of my coffee setup so I can add sticky note annotations when you hover over each item.

For the bookshelves idea specifically, I would love to have a correction system in place. The problem isn't so much SAM as it is Grounding DINO, the model I'm using for object identification. I then pass each identified region to SAM and map the segmentation mask to the box.

Grounding DINO detects a lot of book spines, but often misses 1-2. I am planning to try out YOLO-World (https://github.com/AILab-CVC/YOLO-World), which, in my limited testing, performs better for this task.

pip install -U huggingface_hub umap-learn git+https://github.com/openai/CLIP.git

We will also need two more libraries: OpenAI’s CLIP GitHub repo, enabling us to generate image features with the CLIP model, and the umap-learn library, which will let us apply a dimensionality reduction technique called Uniform Manifold Approximation and Projection (UMAP) to those features to visualize them in 2D:

Biggest missing feature for all these self hosted photo hosting is the lack of a real search. Being able to search for things like "beach at night" is a time saver instead of browsing through hundreds or thousands of photos. There are trained neural networks out there like https://github.com/openai/CLIP which are quite good.

In computer vision, this is known as zero-shot learning, or zero-shot prediction, because the goal is to generate predictions without explicitly being given any example predictions to learn from. With the advent of high quality multimodal models like CLIP and foundation models like Segment Anything, it is now possible to generate remarkably good zero-shot predictions for a variety of computer vision tasks, including:

(Github Repo | Most Popular Model | Paper | Project Page)

the first thing that comes to mind is CLIP: https://github.com/openai/CLIP

Whenever I’m working on semantic search applications that connect images and text, I start with a family of models known as contrastive language image pre-training (CLIP). These models are trained on image-text pairs to generate similar vector representations or embeddings for images and their captions, and dissimilar vectors when images are paired with other text strings. There are multiple CLIP-style models, including OpenCLIP and MetaCLIP, but for simplicity we’ll focus on the original CLIP model from OpenAI. No model is perfect, and at a fundamental level there is no right way to compare images and text, but CLIP certainly provides a good starting point.

in the modal card it says: pretrained text encoders (OpenCLIP-ViT/G and CLIP-ViT/L).

Command: "C:\StabilityMatrix\Packages\stable-diffusion-webui\venv\Scripts\python.exe" -m pip install https://github.com/openai/CLIP/archive/d50d76daa670286dd6cacf3bcd80b5e4823fc8e1.zip --prefer-binary

CLIP might work for your needs.