A combined experimental-numerical approach using digital image correlation (DIC) and finite element simulation in order to get the temperature dependent mechanical behaviour is presented. Results from a series of experiments on a Ti6Al4V titanium alloy sheet are shown. Tensile tests were carried out on specimens along 3 different orientations in order to characterise the material anisotropy. The strain-rates are varied from 10-1 to 10-2 s-1 while observations are made at temperatures from 903 to 1003 K. The samples are heated by Joule effect, which allows to use the image correlation in order to obtain the deformation fields and thus the coefficients of Lankford [1]. Differences in the responses of this alloy are observed in terms of work hardening, strain rate and temperature sensitivities. The Norton-Hoff model and the Hill [2] criterion are used to effectively simulate the observed responses obtained from these experiments. An inverse analysis model using kriging meta-model [3] is applied to determine each parameter of the mechanical behaviour law. The model, with the constants determined from these experiments, is then used to predict the mechanical behaviour of Ti6Al4V. Thus, the model is implemented into the implicit finite element code Forge® to model forming of thin-walled structures. The predictions are found to be very close to the observations.