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 diverse
tasks such as image classification, object detection, and segmentation. This paper presents 4m:21, a unified
multitask system that combines three independently trained models into a single framework. The system
leverages a shared preprocessing pipeline and postprocessing logic to integrate the outputs of the
image classification, object detection, and segmentation models. Experimental results demonstrate that
4m:21 achieves competitive performance on benchmark datasets for all three tasks while maintaining
modularity and flexibility. This work highlights the potential of modular multitask systems for scalable and
efficient computer vision applications