Place recognition plays an essential role in the field of autonomous drivingand robot navigation. Point cloud based methods mainly focus on extractingglobal descriptors from local features of point clouds. Despite having achievedpromising results, existing solutions neglect the following aspects, which maycause performance degradation: (1) huge size difference between objects inoutdoor scenes; (2) moving objects that are unrelated to place recognition; (3)long-range contextual information. We illustrate that the above aspects bringchallenges to extracting discriminative global descriptors. To mitigate theseproblems, we propose a novel method named TransLoc3D, utilizing adaptivereceptive fields with a point-wise reweighting scheme to handle objects ofdifferent sizes while suppressing noises, and an external transformer tocapture long-range feature dependencies. As opposed to existing architectureswhich adopt fixed and limited receptive fields, our method benefits fromsize-adaptive receptive fields as well as global contextual information, andoutperforms current state-of-the-arts with significant improvements on populardatasets.