Radio frequency interference (RFI) is considered as anomalous disruptive parasite signal due to its harmful impact in wireless communication. That is why, RFI mitigation is indispensable to avoid this impact. Detecting and localizing the RFI are the first steps in RFI mitigation process. In this paper, we propose two approaches to detect and localize RFI using the supervised and unsupervised techniques of deep learning. First, our research investigates an object detection algorithm based on convolutionnal neural network as a supervised approach. This proposition is based on the object detection algorithm You Only Look Once v3 (YOLO-v3) trained on real-world data contaminated by multiples sources of RFI. Second, we propose the utilisation of Convolutionnal Autoencoder (CAE) as an unsupervised approach. Experimental results show that the RFI detection by YOLO-v3 is relatively fast and it has an excellent accurate detection rate of 94% and show that the average precision of the YOLO-V3 algorithm can achieve 89%. For CAE, the average precision achieves 78% and outperforms the supervised approach in certain cases.