WorldComp2D: Spatio-semantic Representations of Object Identity and Location from Local Views

Learning latent representations that capture both semantic and spatial information is central to efficient spatio-semantic reasoning. However, many existing approaches rely on implicit latent structures combined with dense feature maps or task-specific heads, limiting computational efficiency and flexibility. We propose WorldComp2D, a novel lightweight representation learning framework that explicitly structures latent space geometry according to object identity and spatial proximity using multiscale \textit{local} receptive fields. This framework consists of (i) a proximity-dependent encoder that maps a given observation into a spatio-semantic latent space and (ii) a localizer that infers the coordinates of objects in the input from the resulting spatio-semantic representation. Using facial landmark localization as a proof-of-concept, we show that, compared to SoTA lightweight models, WorldComp2D reduces the numbers of parameters and FLOPs by up to $4.0\times$ and $2.2\times$, respectively, while maintaining real-time performance on CPU. These results demonstrate that explicitly structured latent spaces provide an efficient and general foundation for spatio-semantic reasoning. This framework is open-sourced at https://github.com/JinSeongmin/WorldComp2D.

For AI to interact with the world, it must understand both "what" an object is and "where" it is located. Most current AI models try to learn this by processing massive amounts of data in a way that is computationally heavy, making them slow and difficult to run on basic devices like standard laptops or smartphones. We developed WorldComp2D, a new, lightweight framework that changes how AI organizes its internal knowledge. Instead of a messy black box of information, WorldComp2D explicitly organizes data like a map, grouping information based on an object's identity and its physical closeness to other things. We tested this by teaching the AI to find facial features, like the corners of eyes or the tip of a nose. Our approach allows the AI to perform complex spatial tasks using up to four times less memory and half the processing power of current state-of-the-art models. This proves that we can make AI much more efficient without sacrificing accuracy. By making these models lighter and faster, we can bring sophisticated spatial intelligence to everyday devices, from better mobile apps to smarter, more responsive robotics.