Current Results

Qualitative Comparisons for Shape Completion

Figure 1. Qualitative results for Shape Completion. Our proposed approach is able to generate diverse plausible 3D shapes consistent with the partial input. The generated shapes are visually consistent with realistic shapes even with significantly missing parts( in Red)

Qualitative Comparisons for Single View Reconstruction

Figure 2. Single-view 3D reconstruction. We visualize multiple shapes predicted by our approach given the input images. We observe meaningful shape variation in the unobserved regions e.g. front of chair in left image.

Qualitative Comparisons for Language based Generations

Figure 3. Language Conditioned Generation. The results signify that our approach can meaningfully estimate the correlation between input description and correspondingly plausible shapes while simultaneously generate the missing context required to generate them.

For more of our amazing results please visit here.