The quantity and spatial distribution of cotton organs reflect the plant's growth responses to soil water, fertilizer, and salinity conditions. Precise organ detection can provide a scientific basis for the regulation of water, fertilizer and salt in mulched drip irrigation. However, cotton organ detection in field environments remains challenging owing to the presence of complex occlusions and the highly variable morphology of target organs, with these difficulties being particularly pronounced in arid regions such as southern Xinjiang. To address these challenges, this study proposes ENT-YOLO, a compact yet high precision detection model built upon the YOLOv11n. First, the CSP-EDLAN module is introduced to replace the original C3k2 structure, effectively reducing model parameters. Second, CIoU loss is integrated with NWD loss to enhance bounding box regression accuracy. Finally, a TADDH detection head is incorporated to enhance feature representation and localization robustness for organs exhibiting variable morphology and blurred texture. Experimental results on the self-constructed cotton organ detection dataset (COD-DS) show that ENT-YOLO achieves 76.36 % precision, 73.44 % recall, and 79.77 % mAP@0.5, with only 8.4 GFLOPs and 2.10 M parameters. The AP for the bud-flower class and the boll class increases by 3.13 % and 1.17 %, respectively, compared with the baseline model. The overall model size is merely 4.2 MB, representing a 19.2 % reduction relative to YOLOv11n. In comparison with mainstream detectors, ENT-YOLO demonstrates an improved balance between accuracy and compactness. Moreover, the spatial distribution map of the target organs is constructed using ENT-YOLO detection outputs, it lays a methodological foundation for the subsequent quantitative analysis of the spatial distribution and quantity of key organs such as buds, flowers and bolls. The results provide useful references for cotton growth stage determination and irrigation management.