DISK: Learning local features with policy gradient

Authors: Michał Tyszkiewicz, Pascal Fua, Eduard Trulls

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our approach on three different benchmarks, and present two ablation studies.
Researcher Affiliation Collaboration Michał J. Tyszkiewicz1 Pascal Fua1 Eduard Trulls2 1École Polytechnique Fédérale de Lausanne (EPFL) 2Google Research, Zurich michal.tyszkiewicz@epfl.ch pascal.fua@epfl.ch trulls@google.com
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Training and inference code is available at https://github.com/cvlab-epfl/disk.
Open Datasets Yes We use a subset of the Mega Depth dataset [19], from which we choose 135 scenes with 63k images in total.
Dataset Splits Yes We rely on a validation set of two scenes: Sacre Coeur and St. Peter s Square .
Hardware Specification No The paper mentions "GPU memory" but does not specify any particular GPU model, CPU, or other hardware specifications used for experiments.
Software Dependencies No The paper mentions using ADAM [17] but does not provide specific version numbers for any software dependencies or libraries.
Experiment Setup Yes Rewards are λtp = 1, λfp = 0.25 and λkp = 0.001. We use a batch of two scenes, with three images in each. We use ADAM [17] with learning rate of 10 4. To pick the best checkpoint, we evaluate performance in terms of pose estimation accuracy in stereo, with DEGENSAC [7]. Specifically, every 5k optimization steps we compute the mean Average Accuracy (m AA) at a 10o error threshold, as in [16]: see Sec. 4.1 and the appendix for details. We anneal λfp and λkp over the first 5 epochs, starting with 0 and linearly increasing to their full value at the end. Grid cells are square, with each side h = 8 pixels.