Convenience class to aply the same operation to a whole set of images.

Random sampling under a constraint: each sample in the batch has the same feature, which is chosen randomly from a known pool of 'features' for

Dataset of outdoor street scenes, 5 images each time

Concatenation of several datasets

Mixture between MLP and DPT head that outputs 3d points and local features (with MLP). The input for both heads is a concatenation of Encoder and

Adapt croco's DPTOutputAdapter implementation for dust3r: remove duplicated weigths, and fix forward for dust3r

Euclidean distance between 3d points

r"""Creates a criterion that measures Learned Perceptual Image Patch Similarity (LPIPS). Arguments: net_type (str): the network type

Linear head for dust3r Each token outputs: - 16x16 3D points (+ confidence)

Optimize a global scene, given a list of pairwise observations. Unlike PointCloudOptimizer, you can fix parts of the optimization process (partia

Artifically augmenting the size of a dataset.

This a Dummy Optimizer. To use only when the goal is to visualize the results for a pair of images (with is_symmetrized)

Implements Adam optimizer with per-point learning rates. Allows unique learning rates for each point in specified parameter tensors, usef

DPT module for dust3r, can return 3D points + confidence for all pixels

Optimize a global scene, given a list of pairwise observations. Graph node: images Graph edges: observations = (pred1, pred2)

Artifically changing the size of a dataset.

Dataset of indoor scenes, 5 images each time

Optimizes a signed distance-function to improve depthmaps.

Dataset of outdoor street scenes, 5 images each time

AP loss.

Virtual Trackball Control. >>> ball = Arcball() >>> ball = Arcball(initial=numpy.identity(4)) >>> ball.place([320, 320], 320) >>> bal

Two siamese encoders, followed by two decoders. The goal is to output 3d points directly, both images in view1's frame (hence the asymmetry).

Optimize a global scene, given a list of pairwise observations. Graph node: images Graph edges: observations = (pred1, pred2)

Define all basic options. Usage: class MyDataset (BaseStereoViewDataset): def _get_views(self, idx, rng): #

Weighted regression by learned confidence. Assuming the input pixel_loss is a pixel-level regression loss. Principle: high-confi

Weight matching by learned confidence. Same as ConfLoss but for a matching criterion Assuming the input matching_loss is a match-level loss.

a dataset that you can easily resize and combine. Examples: --------- 2 * dataset ==> duplicate each element 2x 10 @ dataset

L-norm loss

Handle images with non-square aspect ratio. All images in the same batch have the same aspect ratio. true_shape = [(height, width) ..

Matching loss per image only compare pixels inside an image but not in the whole batch as what would be done usually

Easily combinable losses (also keep track of individual loss values): loss = MyLoss1() + 0.1*MyLoss2() Usage: Inherit from this c

Ensure that all 3D points are correct. Asymmetric loss: view1 is supposed to be the anchor. P1 = RT1 @ D1 P2 = RT2 @ D2

Same than Regr3D but invariant to depth scale. if gt_scale == True: enforce the prediction to take the same scale than GT

Same than Regr3D but invariant to depth shift. if gt_scale == True: enforce the prediction to take the same scale than GT

Same than Regr3D but invariant to depth shift.

Same than Regr3D but invariant to depth shift.