The objective of this work is to achieve sensorless reconstruction of a 3D\nvolume from a set of 2D freehand ultrasound images with deep implicit\nrepresentation. In contrast to the conventional way that represents a 3D volume\nas a discrete voxel grid, we do so by parameterizing it as the zero level-set\nof a continuous function, i.e. implicitly representing the 3D volume as a\nmapping from the spatial coordinates to the corresponding intensity values. Our\nproposed model, termed as ImplicitVol, takes a set of 2D scans and their\nestimated locations in 3D as input, jointly refining the estimated 3D locations\nand learning a full reconstruction of the 3D volume. When testing on real 2D\nultrasound images, novel cross-sectional views that are sampled from\nImplicitVol show significantly better visual quality than those sampled from\nexisting reconstruction approaches, outperforming them by over 30% (NCC and\nSSIM), between the output and ground-truth on the 3D volume testing data. The\ncode will be made publicly available.\n