Leukemia, a common blood malignancy, necessitates fast and accurate detection to provide successful medical intervention. This paper provides a novel method for identifying objects in real-time using two effective deep-learning models: ResNet50 for complex feature extraction, and YOLOv3 for live object recognition A combined approach is created through integration, utterly improving the accuracy of blood cancer cell detection. Preprocessing blood cell microscopic pictures to enhance their quality and facilitate dimensionality reduction is part of the procedure, these improved images are then sent into ResNet50 to acquire key features, which are then used to train YOLOv3 to recognize objects. The integrated model is rigorously trained on an extensive set of analysed leukemia photos to ensure strong performance. Extensive tests on several leukemia forms and imaging settings confirm our combined model's far better specificity (Xu et al. in IEEE Trans Intell Transp Syst 23:19760–19771, 2022), sensitivity, and accuracy than previous approaches. Because of YOLOv3’s instantaneous processing capabilities, our technique is appropriate for speedy and mechanized leukemia assessment in clinical settings. This hybrid approach not only enhances diagnostic imaging but possesses the potential to change leukemia examination, leading to better patient outcomes and increased healthcare efficiency (Abunadi and Senan in Sensors 24:1629, 2022).