Feature/dataset

[aiorhiroki]

dataset, dataloaderの実装

[yumion]

augmentationとpreprocessingを明確に分けたいです(image_utilがpreprocessingを含んでいるので機能を分けたい...) こんな感じに

        # apply augmentations
        if self.augmentation:
            sample = self.augmentation(image=image, mask=mask)
            image, mask = sample['image'], sample['mask']

        # apply preprocessing
        if self.preprocessing:
            sample = self.preprocessing(image=image, mask=mask)
            image, mask = sample['image'], sample['mask']

[yumion]

callbacksの GenerateSampleResult 、 IoUHistory をデータのlistじゃなくてgenerator(dataloader)にしたいです そうするとこういうの

    for sample_image_path in annotations:
        input_image_path, mask_image_path = sample_image_path
        sample_image = image_util.read_image(
            input_image_path, anti_alias=True)
        segmented = image_util.read_image(
            mask_image_path, normalization=False, train_colors=train_colors)

が毎回書かれずにスッキリします

[yumion]

classificatoin (image, video両方)、segmentation w/ data augmentation、segmentationのoptunaが動くことを確認しました

[yumion]

classification(image,video),segmentation(aug, optuna)で動くことを確認しました〜