Ref: https://scikit-learn.org/stable/auto_examples/model_selection/ plot_confusion_matrix.html#sphx-glr-auto-examples-model-selec

Print kernel information. This is used by prof.py.

----------------------------------------------------------------------------

Args: - filename: string representing - label_from: string representing "source" taxonomy name - label_to: string representing "sink"/ "targe

Args: - Returns: -

Removes keypoints too close to the border

Args: Returns: - y_true - y_pred

Removes the given reparameterization of a parameter from a module. If no parameter is supplied then all reparameterizations are removed.

Args: - Returns: - None

Since overlaid class text is difficult to read below 1080p, we upsample predictions.

Creates Parameters to be used for reparameterization and creates names that for attributes for the module these Parameters will corre

Checks if a string s represents an int. Args: - s: string Returns: - boolean, whether string represents an integer value

Call this method to disable sparsity and restore all weights. This will only work if init(...) was called with allow_recompute=True.

We ensure that a 1:1 mapping is even possible (unique keys and values). The following would fail: {'a': 1, 'b': 2, 'c': 3, 'd': 3}

Revert effect of previously calling partial_step.

Revert effect of previously calling partial_step.

Interpolate descriptors at keypoint locations

----------------------------------------------------------------------------

Save a Python dictionary to a JSON file. Args: json_fpath: Path to file to create. dictionary: Python dictionary to be serialized.

Set direction (fprop, bprop) based on PyTorch sequence markers. It is a heuristic and not a foolproof method.

Detect and set the class/module (mod) and operation (op) of the kernel e.g. torch.nn.functional / linear, torch / sigmoid. The lookup sequence

Fast Non-maximum suppression to remove nearby points

Returns a dict containing the current state of this :class:`FP16_Optimizer` instance. This dict contains attributes of :class:`FP16_O

If no closure is supplied, :attr:`step` should be called after ``fp16_optimizer_obj.backward(loss)``. :attr:`step` updates t

Strided check for overflow. You can get status by calling has_overflow.

Strided check for overflow. You can get status by calling has_overflow.

Make sure that mapping on disk is correct.

Person vs. Motorcyclist in center Relabeled model correctly predicts `motorcylist`. for `motorcylist`. Motorcyclist silhouette patte

Person vs. Motorcyclist in center. Relabeled model incorrectly predicts `person` instead of `motorcylist`. [0,0,0,0],

When labels were inaccurate, we often marked them as `unlabeled`, e.g. COCO cabinets included `counter` pixels.

Ensure oracle model testing script works correctly. base_sizes=( #360 720 #1080 python -u mseg_semantic/tool/test_or

Ensure universal model testing script works correctly. base_sizes=( #360 720 #1080 python -u mseg_semantic/tool/test

Compare w/ speed of set approach

Ensure we can find the classes to exclude when evaluating a "relabeled" MSeg model on the val split of a training dataset.

Get a specific mask out of an image, and compare a few summary statistics to make sure correct mask is obtained.