Source code for mmagic.evaluation.metrics.equivariance
# Copyright (c) OpenMMLab. All rights reserved.
from collections import defaultdict
from copy import deepcopy
from typing import Iterator, List, Optional, Sequence
import numpy as np
import torch
import torch.nn as nn
from mmengine.dist import all_gather
from torch.utils.data.dataloader import DataLoader
from mmagic.registry import METRICS
from .base_gen_metric import GenerativeMetric
@METRICS.register_module('EQ')
@METRICS.register_module()
[docs]class Equivariance(GenerativeMetric):
def __init__(self,
fake_nums: int,
real_nums: int = 0,
fake_key: Optional[str] = None,
real_key: Optional[str] = 'gt_img',
need_cond_input: bool = False,
sample_mode: str = 'ema',
sample_kwargs: dict = dict(),
collect_device: str = 'cpu',
prefix: Optional[str] = None,
eq_cfg=dict()):
super().__init__(fake_nums, real_nums, fake_key, real_key,
need_cond_input, sample_mode, collect_device, prefix)
# set default sampler config
self._eq_cfg = deepcopy(eq_cfg)
self._eq_cfg.setdefault('compute_eqt_int', False)
self._eq_cfg.setdefault('compute_eqt_frac', False)
self._eq_cfg.setdefault('compute_eqr', False)
self._eq_cfg.setdefault('translate_max', 0.125)
self._eq_cfg.setdefault('rotate_max', 1)
self.SAMPLER_MODE = 'EqSampler'
self.sample_kwargs = sample_kwargs
# compute numbers of eq
self.n_sub_metric = 0
if self._eq_cfg['compute_eqt_int']:
self.n_sub_metric += 1
if self._eq_cfg['compute_eqt_frac']:
self.n_sub_metric += 1
if self._eq_cfg['compute_eqr']:
self.n_sub_metric += 1
self.fake_results = defaultdict(list)
@torch.no_grad()
[docs] def process(self, data_batch: dict, data_samples: Sequence[dict]) -> None:
"""Process one batch of data samples and predictions. The processed
results should be stored in ``self.fake_results``, which will be used
to compute the metrics when all batches have been processed.
Args:
data_batch (dict): A batch of data from the dataloader.
data_samples (Sequence[dict]): A batch of outputs from the model.
"""
cfg_key_list = ['compute_eqt_int', 'compute_eqt_frac', 'compute_eqr']
sample_key_list = ['eqt_int', 'eqt_frac', 'eqr']
for pred in data_samples:
for cfg_key, sample_key in zip(cfg_key_list, sample_key_list):
if self._eq_cfg[cfg_key]:
assert sample_key in pred
# assert hasattr(pred, sample_key)
eq_sample = pred[sample_key]
diff = eq_sample['diff'].to(torch.float64).sum()
mask = eq_sample['mask'].to(torch.float64).sum()
self.fake_results[sample_key] += [diff, mask]
[docs] def get_metric_sampler(self, model: nn.Module, dataloader: DataLoader,
metrics: List[GenerativeMetric]):
"""Get sampler for generative metrics. Returns a dummy iterator, whose
return value of each iteration is a dict containing batch size and
sample mode to generate images.
Args:
model (nn.Module): Model to evaluate.
dataloader (DataLoader): Dataloader for real images. Used to get
batch size during generate fake images.
metrics (List['GenerativeMetric']): Metrics with the same sampler
mode.
Returns:
:class:`dummy_iterator`: Sampler for generative metrics.
"""
batch_size = dataloader.batch_size
sample_model = metrics[0].sample_model
assert all([metric.sample_model == sample_model for metric in metrics
]), ('\'sample_model\' between metrics is inconsistency.')
return eq_iterator(
batch_size=batch_size,
max_length=max([metric.fake_nums_per_device
for metric in metrics]),
sample_mode=sample_model,
eq_cfg=self._eq_cfg,
sample_kwargs=self.sample_kwargs)
[docs] def compute_metrics(self, results) -> dict:
"""Compute the metrics from processed results.
Args:
results (list): The processed results of each batch.
Returns:
dict: The computed metrics. The keys are the names of the metrics,
and the values are corresponding results.
"""
results = dict()
for key in ['eqt_int', 'eqt_frac', 'eqr']:
if key not in self.fake_results:
continue
sums = torch.stack(self.fake_results[key], dim=0)
mses = (sums[0::2] / sums[1::2]).mean()
psnrs = np.log10(2) * 20 - mses.log10() * 10
psnrs = psnrs.cpu().numpy()
results[key] = psnrs
return results
[docs] def _collect_target_results(self, target: str) -> Optional[list]:
"""Collect function for Eq metric. This function support collect
results typing as Dict[List[Tensor]]`.
Args:
target (str): Target results to collect.
Returns:
Optional[list]: The collected results.
"""
if target == 'real':
return
results = getattr(self, f'{target}_results')
results_collected = []
results_collected = dict()
for key, result in results.items():
result_collected = torch.stack(result)
result_collected = torch.cat(all_gather(result_collected), dim=0)
results_collected[key] = torch.split(result_collected,
len(result_collected))
return results_collected
[docs]class eq_iterator:
def __init__(self, batch_size, max_length, sample_mode, eq_cfg,
sample_kwargs) -> None:
self.batch_size = batch_size
self.max_length = max_length
self.sample_mode = sample_mode
self.eq_cfg = deepcopy(eq_cfg)
self.sample_kwargs = sample_kwargs
[docs] def __next__(self) -> dict:
if self.idx >= self.max_length:
raise StopIteration
self.idx += self.batch_size
mode = dict(
sample_mode=self.sample_mode,
eq_cfg=self.eq_cfg,
sample_kwargs=self.sample_kwargs)
# StyleGAN3 forward will receive eq config from mode
return dict(inputs=dict(mode=mode, num_batches=self.batch_size))