Diffractions play a significant role in seismic processing and imaging since they can image structures smaller than the seismic wavelength, such as discontinuities, faults, and pinch-outs. The traveltime of a non-migrated stacked diffraction event typically has a hyperbolic shape around its apex, which collapses after a migration procedure. We can interpret such apex as the time image of a point diffractor. An essential problem in diffraction analysis is the detection of those apexes in the generally noisy data environment because their position and processing parameters (such as migration velocity) play an essential role in obtaining more reliable and accurate imaging results. In this work, we introduce a Fully Convolutional Network (namely, LeNet-5 FCN) to automatic detect diffraction apexes on real seismic data. To deal with the low amount of annotated data, we propose to use data augmentation (e.g., polarity inversion, automatic gain control, zoom) and ensemble strategies. By combining our LeNet-5 FCN with those strategies, we reached 91.2% average accuracy on three land seismic datasets.