Differentiable Product Quantization for Memory Efficient Camera Relocalization
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866929433213927424 |
|---|---|
| author | Laskar, Zakaria Melekhov, Iaroslav Benbihi, Assia Wang, Shuzhe Kannala, Juho |
| author_facet | Laskar, Zakaria Melekhov, Iaroslav Benbihi, Assia Wang, Shuzhe Kannala, Juho |
| contents | Camera relocalization relies on 3D models of the scene with a large memory footprint that is incompatible with the memory budget of several applications. One solution to reduce the scene memory size is map compression by removing certain 3D points and descriptor quantization. This achieves high compression but leads to performance drop due to information loss. To address the memory performance trade-off, we train a light-weight scene-specific auto-encoder network that performs descriptor quantization-dequantization in an end-to-end differentiable manner updating both product quantization centroids and network parameters through back-propagation. In addition to optimizing the network for descriptor reconstruction, we encourage it to preserve the descriptor-matching performance with margin-based metric loss functions. Results show that for a local descriptor memory of only 1MB, the synergistic combination of the proposed network and map compression achieves the best performance on the Aachen Day-Night compared to existing compression methods. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2407_15540 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Differentiable Product Quantization for Memory Efficient Camera Relocalization Laskar, Zakaria Melekhov, Iaroslav Benbihi, Assia Wang, Shuzhe Kannala, Juho Computer Vision and Pattern Recognition Camera relocalization relies on 3D models of the scene with a large memory footprint that is incompatible with the memory budget of several applications. One solution to reduce the scene memory size is map compression by removing certain 3D points and descriptor quantization. This achieves high compression but leads to performance drop due to information loss. To address the memory performance trade-off, we train a light-weight scene-specific auto-encoder network that performs descriptor quantization-dequantization in an end-to-end differentiable manner updating both product quantization centroids and network parameters through back-propagation. In addition to optimizing the network for descriptor reconstruction, we encourage it to preserve the descriptor-matching performance with margin-based metric loss functions. Results show that for a local descriptor memory of only 1MB, the synergistic combination of the proposed network and map compression achieves the best performance on the Aachen Day-Night compared to existing compression methods. |
| title | Differentiable Product Quantization for Memory Efficient Camera Relocalization |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2407.15540 |