Same as add just more pythonic.

Copies this config into a new config object, making the changes given by new_config_dict.

Copies new_config_dict into this config object. Note: new_config_dict can also be a config object.

Returns the average of the elements in the window.

Sets the active config. Works even if cfg is already imported!

Adds an element to the window, removing the earliest element if necessary.

Here one layer means a string of n Bottleneck blocks.

Loads weights from a compressed save file.

Resets the MovingAverage to its initial state.

Adapted from https://discuss.pytorch.org/t/how-to-train-with-frozen-batchnorm/12106/8

Note: config should be config.name.

Initialize weights for training.

instances: Dict {'masks': [N x H x W], torch.uint8, 'boxes': [N x 4], torch.float64, 'classe

Saves the model's weights using compression because the file sizes were getting too big.

Sets the dataset of the current config.

Form 3D points groups, with 3 points in each grouup. :param p123: points index :param pw: 3D points :return:

Form 3D points groups, with 3 points in each grouup. :param p123: points index :param pw: 3D points :return:

Resize image to a given size. Args: img (ndarray): The input image. size (tuple[int]): Target size (w, h). return_scale (b

Start timing the specific function. Note: If use_stack is True, only one timer can be active at a time. Once you stop this timer, the previo

If use_stack is True, this will stop the currently running timer and restore the previous timer on the stack if that exists. Note if use_stack is T

Rescale a size by a ratio. Args: size (tuple[int]): (w, h). scale (float): Scaling factor. Returns: tuple[int]: scaled

Call this once per image.

Add a downsample layer to the backbone as per what SSD does.

Compute the jaccard overlap of two sets of boxes. The jaccard overlap is simply the intersection over union of two boxes. Here we operate on

initialize conv/fc bias value according to giving probablity

Computation of error metrics between predicted and ground truth depths. Prediction and ground turth need to be converted to the same unit e.g

Compute gradient map from depth map with 3x3 sobel filter

Constructs a backbone given a backbone config object (see config.py).

Ensures that the image is the given size without distorting aspect ratio.

Warning: result not cached.

Resize image while keeping the aspect ratio. Args: img (ndarray): The input image. scale (float | tuple[int]): The scaling factor

Initializes the backbone weights for training.

Initialize the console to be able to use ANSI escape characters on Windows.

We resize both tensors to [A,B,2] without new malloc: [A,2] -> [A,1,2] -> [A,B,2] [B,2] -> [1,B,2] -> [A,B,2] Then we compute the area of

Write losses to the event file

Write visaul examples to the event file

Computes the pariwise mask IoU between two sets of masks of size [a, h, w] and [b, h, w]. The output is of size [a, b].

Returns the original image object at index in OPENCV form (BGR) Argument: index (int): index of img to show Return:

Calculate the new size to be rescaled to. Args: old_size (tuple[int]): The old size (w, h) of image. scale (float | tuple[int]): T