Expensive scientific camera hardware is amongst the main cost factors in modern, high-performance microscopes. On the other hand, cheap, consumer-grade camera devices can provide surprisingly good performance. Widely available smartphones include cameras, providing a good opportunity for "imaging on a budget". Yet, Single-Molecule-Localization-Microscopy (SMLM) techniques like Photoactivated Localization Microscopy (PALM) or (direct) Stochastic Optical Reconstruction Microscopy dSTORM, are demanding in terms of photon sensitivity and readout noise, seemingly requiring a scientific-grade camera. Here we show that super-resolution imaging by dSTORM is possible using a consumer grade cellphone camera. Trained image-to-image generative adversarial network (GAN), successfully improves the signal-to-noise ratio (SNR) by compensating noise and compression artifacts in the acquired video-stream at poor imaging conditions. We believe that "cellSTORM" paves the way for affordable super-resolution microscopy suitable for research and education. Our low-cost setup achieves optical resolution below 80,nm yielding wide access to cutting edge research to a big community.