4M:21: A UNIFIED MULTITASK MODEL FOR IMAGE CLASSIFICATION, OBJECT DETECTION, AND SEGMENTATION

Multitask systems in computer vision often require the integration of multiple models to perform diversetasks such as image classification, object detection, and segmentation. This paper presents 4m:21, a unifiedmultitask system that combines three independently trained models into a single framework. The systemleverages a shared preprocessing pipeline and postprocessing logic to integrate the outputs of theimage classification, object detection, and segmentation models. Experimental results demonstrate that4m:21 achieves competitive performance on benchmark datasets for all three tasks while maintainingmodularity and flexibility. This work highlights the potential of modular multitask systems for scalable andefficient computer vision applications