A callable which takes a dataset dict in Scannet Dataset format, and map it into a format used by the model.

A callable which takes a dataset dict in Detectron2 Dataset format, and map it into a format used by MaskFormer. This dataset mapper app

Params for coco evaluation api

Absolute pos embedding, learned.

A basic Swin Transformer layer for one stage. Args: dim (int): Number of feature channels depth (int): Depths of this stage.

The standard ResNet stem (layers before the first residual block), with a conv, relu and max_pool.

Evaluate AR for object proposals, AP for instance detection/segmentation, AP for keypoint detection outputs using COCO's metrics. See htt

Modified version of COCOeval for evaluating AP with a custom maxDets (by default for COCO, maxDets is 100)

Wraps another sampler to yield a mini-batch of indices. It enforces that the batch only contain elements from the same group. It also tri

Very simple multi-layer perceptron (also called FFN)

Multilayer perceptron.

In training, we only care about the "infinite stream" of training data. So this sampler produces an infinite stream of indices and all wo

This class computes an assignment between the targets and the predictions of the network For efficiency reasons, the targets don't include the no

This class computes the loss for DETR. The process happens in two steps: 1) we compute hungarian assignment between ground truth boxes and

Image to Patch Embedding Args: patch_size (int): Patch token size. Default: 4. in_chans (int): Number of input image channels. Def

Absolute pos embedding, learned.

This is a more standard version of the position embedding, very similar to the one used by the Attention is all you need paper, generalized t

This is a more standard version of the position embedding, very similar to the one used by the Attention is all you need paper, generalized t

Implement :paper:`ResNet`.

Swin Transformer Block. Args: dim (int): Number of input channels. num_heads (int): Number of attention heads. window_size

Extension of the Trainer class adapted to MaskFormer.

Window based multi-head self attention (W-MSA) module with relative position bias. It supports both of shifted and non-shifted window. Args:

Ball Query. Find nearby points in spherical space.

The basic residual block for ResNet-18 and ResNet-34 defined in :paper:`ResNet`, with two 3x3 conv layers and a projection shortcut if needed

Used for nn.Transformer that uses a HW x N x C rep

The standard bottleneck residual block used by ResNet-50, 101 and 152 defined in :paper:`ResNet`. It contains 3 conv layers with kernels

This is a slightly modified version of the original COCO API, where the functions evaluateImg() and accumulate() are implemented in C++ to sp

Similar to :class:`BottleneckBlock`, but with :paper:`deformable conv <deformconv>` in the 3x3 convolution.

Main class for mask classification semantic segmentation architectures.

Patch Merging Layer Args: dim (int): Number of input channels. norm_layer (nn.Module, optional): Normalization layer. Default: nn

Random Ball Query. Find nearby points in spherical space.