Load frame paths and annotations from files Args: cfg (CfgNode): config

Padding all the tensors from different GPUs to the largest ones. Args: tensor (tensor): tensor to pad. group (group): pytorch

Returns a 1D tensor that contains the indices of instances that have wrong predictions, where true labels in in the specified subset.

Register hooks to layers in `self.layers_names`.

Register hooks to layers in `self.target_layers`.

Register hook to a layer, given layer_name, to obtain activations. Args: layer_name (str): name of the layer.

Register forward and backward hook to a layer, given layer_name, to obtain gradients and activations. Args: layer

Removes entries with invalid boxes. A box is invalid if either its xmax is smaller than its xmin, or its ymax is smaller than its ymi

Round number of layers based on depth multiplier.

Round width of filters based on width multiplier Args: width (int): the channel dimensions of the input. mult

Seriialize the tensor to ByteTensor. Note that only `gloo` and `nccl` backend is supported. Args: data (data): data to be ser

Suppresses printing from the current process.

Update grouth truth statitistics. 1. Difficult boxes are ignored when counting the number of ground truth instances as done in Pascal

Checks if a label map is valid. Args: label_map: StringIntLabelMap to validate. Raises: ValueError: if label map is invalid.

Add the corresponding action predictions.

Visualize and add tensors of n-dimentionals to a Tensorboard SummaryWriter. Tensors will be visualized as a 2D grid image. Args:

Add input to tensorboard SummaryWriter as a video. Args: vid_tensor (tensor): shape of (B, T, C, H, W). Values should lie

Join a list of label list. Args: labels (list): The input label list. Returns: labels (list): The joint list of all lists

Synchronize running_mean, and running_var. Call this before eval.

Computes area of boxes. Args: boxlist: BoxList holding N boxes Returns: a numpy array with shape [N*1] representing box areas

Perform color brightness on the given image. Args: var (float): variance. image (array): image to perform color brightness.

Perfrom brightness jittering on the input images. The channels of images should be in order BGR. Args: var (float): jitter ratio

Perform color brightness on the given list of images. Args: var (float): variance. images (list): list of images to perform c

Build a dataset, defined by `dataset_name`. Args: dataset_name (str): the name of the dataset to be constructed. cfg (CfgNode

Build training model and its associated tools, including optimizer, dataloaders and meters. Args: cfg (CfgNode): configs. Details

Convert BN parameters to Sub-BN parameters if model contains Sub-BNs. Args: key (OrderedDict): source dict of parameters. mdo

Convert BN parameters to Sub-BN parameters if model contains Sub-BNs. Args: key (OrderedDict): source dict of parameters. mdo

Decomposed Spatial Relative Positional Embeddings.

Temporal Relative Positional Embeddings.

Update the stats in bn layers by calculate the precise stats. Args: loader (loader): data loader to provide training data. mo

Clean up open video files and windows.

Close Tensorboard writer.

Perfrom a color jittering on the input images. The channels of images should be in order BGR. Args: images (tensor): images to pe

Evaluates detections as being tp, fp or ignored from a single image. The evaluation is done in two stages: 1. All detections are mat

Concatenate list of BoxLists. This op concatenates a list of input BoxLists into a larger BoxList. It also handles concatenation of BoxList

Perform color contrast on the given image. Args: var (float): variance. image (array): image to perform color contrast. R

Perfrom contrast jittering on the input images. The channels of images should be in order BGR. Args: var (float): jitter ratio fo

Perform color contrast on the given list of images. Args: var (float): variance. images (list): list of images to perform col

Loads label map proto and returns categories list compatible with eval. This function loads a label map and returns a list of dicts, each of whic

Creates dictionary of COCO compatible categories keyed by category id. Args: categories: a list of dicts, each of which has the following k

Crop the boxes given the offsets. Args: boxes (array): boxes to crop. x_offset (int): offset on x. y_offset (int): of

Either display a single frame (BGR image) to a window or write to an output file if output path is provided. Args:

Draw a list of text labels for some bounding box on the image in downward direction. The next text label will be below the previous o

Draw a list of text labels for some bounding box on the image in upward direction. The next text label will be on top of the previous

Draw labels and bouding boxes for one image. By default, predicted labels are drawn in the top left corner of the image or correspond

Draw prediction for the given task. Args: task (TaskInfo object): task object that contain the necessary information for

Draw predicted and ground-truth (if provided) results on the video/folder of images. Write the visualized result to a video output fi

Start to record time.

Stop to record time.

Evaluate the model on the val set. Args: val_loader (loader): data loader to provide validation data. model (model): model to

Run AVA evaluation given numpy arrays.

Horizontally flip the boxes. Args: boxes (array): box to flip. im_width (int): width of the image. Returns: boxes

Return a task object in the correct order based on task id if result(s) is available. Otherwise, raise queue.Empty exception.

Convert our data format into the data format used in official AVA evaluation.

Get the groundtruth annotations corresponding the "subset" of AVA val set. We define the subset to be the frames such that (second % 4 == 0).

Get corner coordinates of boxes. Returns: a list of 4 1-d numpy arrays [y_min, x_min, y_max, x_max]

Given a schedule and epoch index, return the long cycle base shape. Args: schedule (configs): configs that contains training and mult

Generates the box coordinates for cutmix. Args: img_shape (tuple): Image shape as tuple lam (float): Cutmix lambda value

Parse numpy-like fancy indexing from a string. Args: string (str): string represent the indices to take a subset of from

Get input clip from the video/folder of images for a given keyframe index. Args: keyframe_idx (int): index of the

Retrieve the logger with the specified name or, if name is None, return a logger which is the root logger of the hierarchy. Args:

Based on multigrid hyperparameters, define the schedule of a long cycle. Args: cfg (configs): configs that contains train

Return a video writer object. Args: path (str): path to the output video file. fps (int or float): frames per

Return a video writer object. Args: path (str): path to the output video file. fps (int or float): frames per

Return a video writer object. Args: path (str): path to the output video file.

Get parameters for ``crop`` for a random sized crop. Args: img (PIL Image): Image to be cropped. scale (tuple): range

Get the full path to a checkpoint file. Args: path_to_job (string): the path to the folder of the current job. epoch (int): t

Get the full path to a checkpoint file. Args: path_to_job (string): the path to the folder of the current job. epoch (int): t

Predict and append prediction results to each box in each keyframe in `self.pred_boxes` dictionary.

Get the rank of the current process.

Given the video index, return the list of sampled frame indexes. Args: index (int): the video index. Returns:

Given the video index, return the list of indexs of sampled frames. Args: index (int): the video index. Returns:

Sample frames among the corresponding clip. Args: center_idx (int): center frame idx for current clip half_len (int): half o

Retrieves the stem module by name.

Retrieves the lr step index for the given epoch. Args: cfg (CfgNode): configs. Details can be found in slowfast/config/de

Retrieves the transformation module by name.

Returns weights from registered layers. Returns: weights (Python dictionary): a dictionary of the pair {layer

Compute the GPU memory usage for the current device (GB).

Inflate 2D model weights in state_dict_2d to the 3D model weights in state_dict_3d. The details can be found in: Joao Carreira, and Andre

Inflate 2D model weights in state_dict_2d to the 3D model weights in state_dict_3d. The details can be found in: Joao Carreira, and Andre

Update cfg based on multigrid settings. Args: cfg (configs): configs that contains training and multigrid specific

Computes pairwise intersection-over-area between box and mask collections. Intersection-over-area (ioa) between two masks mask1 and mask2 is defi