<p>Conventional approaches to femtosecond (fs) pulse characterization depend on nonlinearities without exception, thereby impeding their utilization in measuring weak fs pulses. Their reliance on iterative phase retrieval or spectrometer-based readout further prevents single-shot full-field characterization over high-repetition-rate fs pulse trains. Here, we present linear spectral shearing interferometry (LSSI) based intelligent single-shot full-field characterization (ISFC) towards high-repetition-rate fs pulse trains, where both intensity and phase of fs pulses are reconstructed in single shot from temporal interferograms through a well-trained fully-connected neural network. The spectral shear in LSSI is created completely by linear effects enabling measurement of weak fs pulses. We validate the proposed method on a megahertz fs pulse train with picojoule-level energy. Moreover, the switching dynamics of a programmable spectral filter are successfully resolved with the proposed method at a megahertz frame rate. We anticipate LSSI-based ISFC to be a particularly powerful tool for characterizing weak ultraviolet fs pulses, and even attosecond pulses with high repetition rates.</p>