r""" Parameters ---------- grad_out : torch.Tensor (B, C, npoint, nsample) tensor of the gradients of the output

Compute 2D bounding box IoU. Input: box1: tuple of (xmin,ymin,xmax,ymax) box2: tuple of (xmin,ymin,xmax,ymax) Output:

Compute 3D bounding box IoU. Note: only for axis-aligned bounding boxes Input: corners1: numpy array (N,8,3), assume up directio

corners: (8,3) no assumption on axis direction

same output as PIL.ImageEnhance.Contrast

Build a dictionary of datasets, keyed by split 'train', 'valid', 'test'. Download dataset and annotations automatically if not exist.

This implementation downloads the remote resource and caches it locally. The resource will only be downloaded if not previously requested.

Utility for deleting a directory. Useful for cleaning the storage space that contains various training artifacts like checkpoints, data etc.

:param batch: :return: coords: N x 4 (batch,x,y,z) feats: N x 3 labels: N colors: B x C x

:param batch: :return: coords_batch: N x 4 (x,y,z,batch)

:param batch: :return: coords: N x 4 (batch,x,y,z) feats: N x 3 labels: N colors: B x C x

:param batch: :return: coords_batch: N x 4 (x,y,z,batch)

same output as PIL.ImageEnhance.Color

Use accumulated predictions and groundtruths to compute Average Precision.

Computes correspondances of points to pixels :param points: tensor containing all points of the point cloud (num_points, 3)

same output as PIL.ImageEnhance.Contrast

Get image color palette for visualizing masks

Compute area of two convex hull's intersection area. p1,p2 are a list of (x,y) tuples of hull vertices. return a list of (x,y) for th

Simply create the symlinks for a given file1 to file2. Useful during model checkpointing to symlinks to the latest successful checkpoint.

A property to get and create dataloaders by split just in need. If no train_dataset_ratio is provided, concatenate map-style dataset

A property to get and create dataloaders by split just in need. If no train_dataset_ratio is provided, concatenate map-style dataset

Overwrite model.train with this function to make sure train/eval mode does not change anymore.

Download a file from google drive Downloading an URL from google drive requires confirmation when the file of the size is too big (google

same output as PIL.ImageOps.equalize PIL's implementation is different from cv2.equalize

Generic functions to compute precision/recall for object detection for multiple classes. Input: pred_all: map of {img_id:

Set to True to skip training.

extract CLIP image features from a folder of images.

Extract features for multimodal or unimodal samples. Args: samples (dict): A dictionary of samples, containing the follow

Flip X-right,Y-forward,Z-up to X-right,Y-down,Z-forward Input and output are both (N,3) array

Forward method.

encode the input descriptions

encode the input descriptions