Person re-identification (re-ID) aims to identify the same person of interest across non-overlapping capturing cameras, which plays an important role in visual surveillance applications and computer vision research areas. Fitting a robust appearance-based representation extractor with limited collected training data is crucial for person re-ID due to the high expanse of annotating the identity of unlabeled data. In this work, we propose a Stronger Baseline for person re-ID, an enhancement version of the current prevailing method, namely, Strong Baseline, with tiny modifications but a faster convergence rate and higher recognition performance. With the aid of Stronger Baseline, we obtained the third place (i.e., 0.94 in mAP) in 2021 VIPriors Re-identification Challenge without the auxiliary of ImageNet-based pre-trained parameter initialization and any extra supplemental dataset.