Sergio Izquierdo

Hi, you can download it from [here](https://www.di.ens.fr/willow/research/netvlad/), in **Downloads** -> **Additional Data** -> **All dataset specifications**.

I've just pushed a new version which includes MSLS_Test. Now you can run the eval.py code with MSLS_Test and a .txt file with the predictions will be generated. You can...

Hi @whulc , You should create a dataset class for your images, like [kitti.py](https://github.com/serizba/SfM-TTR/blob/main/sfmttr/data/kitti.py). Then you can create a [TunerDataset](https://github.com/serizba/SfM-TTR/blob/main/sfmttr/tuner_data.py) using your dataset and the path to the reconstructed COLMAP...

Hi! No, ground truth depth map is only used for evaluation and is not used at all during the finetuning. You can just return a dummy placeholder if you don't...

I've just pushed a new version which includes MSLS_Test. Now you can run the eval.py code with MSLS_Test and a .txt file with the predictions will be generated. You can...

Hi @chenjack9017 Sorry for the late reply. I have updated the README with ckpts of smaller descriptors. Check it [here](https://github.com/serizba/salad/tree/main?tab=readme-ov-file#evaluation).

Hi @qyangcv I have updated the README with links to the ckpts with smaller descriptors. Available [here](https://github.com/serizba/salad/tree/main?tab=readme-ov-file#evaluation).

This looks great, thanks @sarlinpe !

Hi @chennuo0125-HIT The `N` features from DINOv2 are actually the `H // 14 * W // 14`. For every patch of the input image DINOv2 returns a `C` vector, so...

Hi @BobH62 No, we have not considered that but it sounds like a promising future direction. If I remember correctly there are already a few papers doing LiDAR based Visual...