A fundamental challenge in computer vision is object detection; however, it is still difficult to recognize small things because of their limited pixel representation, size changes, and background noise. In order to tackle these problems, we present TRYOLO, an improved real-time object identification model designed primarily for small object recognition and based on YOLOv11. Two important architectural improvements are included in our model: the DeepFocus Block, which enhances spatial feature extraction by adding more convolutions and residual connections, and the C3XFormer Block, which uses positional embeddings and multi-head self-attention to capture global dependencies and improve contextual understanding. These improvements significantly enhance the accuracy of feature representation and detection, especially for small objects in complicated situations. TRYOLO delivers state-of-the-art performance when we test it on two benchmark datasets, VisDrone-DET2019 and GlobalWheat2020. The validation performance metric shows a +2.4% gain in mAP on VisDrone-DET2019 and a +1.4% improvement in mAP on GlobalWheat2020.