Super-image

Updated
- Update `EdsrModel` to fix `no_upsampling` bug where `self.args` was not stored.

Added
- New model: `PhysicssrModel`
- New model: `DrnModel`
- New model: `HanModel`
- New model: `AwsrnModel`

Added
- `calculate_mean_std(dataset)` to `~super_image.utils.metrics` for calculating RGB pixel mean and standard deviation over a dataset.
- New model: `DrlnModel`
- New model: `RcanModel`
- New model: `MdsrModel`
- New model: `CarnModel`
- New model: `PanModel`

Updated
- Update `EdsrModel` to include `no_upsampling` option so it can be reused for `JiifModel`.

Added
- `TrainDataset(dataset)` with support for huggingface datasets.
- `augment_five_crop` function for use with `dataset.map(augment_five_crop, batched=True)`

Added
- Added metrics `EvalMetrics().evaluate(model, eval_dataset)` class to calculate PSNR and SSIM on a model and
evaluation dataset. Accepts `EvalDataset(dataset)` with huggingface datasets.

Changed
- Replaced `EvalDataset(dataset)` to use huggingface datasets instead of H5 files.
- Fixed `EvalDataset(dataset)` to be robust to wrongly sized HR images (not equals to scaled LR image size).