To enable robotic weed control, we develop algorithms to detect nutsedge weed from Bermudagrass turf. Due to the similarity between the weed and the background turf, it is expensive and error-prone to perform manual data labeling. Consequently, directly applying deep learning methods for object detection cannot generate satisfactory results. Building on an instance detection approach, (i.e. Mask R-CNN), we combine synthetic data with raw data to train the network. We propose an algorithm to generate high fidelity synthetic data, adopting different levels of annotations to reduce labeling cost. Moreover, we construct a nutsedge skeleton-based probabilistic map (NSPM) as the neural network input to reduce the reliance on pixel-wise precise labeling. We also modify loss function from cross entropy to Kullback–Leibler divergence which accommodates uncertainty in the labeling process. We have implemented the proposed algorithm and compare it with Faster R-CNN, a typical object detection approach. The results show that our design can effectively reduce the impact of imprecise and insufficient training sample issues and significantly outperforms the counterpart with a false negative rate of 0.4%, a satisfying result for weed control applications.