Skip to content

DSSWAT-5474_add_custom_transforms_on_the_fly #2

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: trainer-baseline
Choose a base branch
from
Open

DSSWAT-5474_add_custom_transforms_on_the_fly #2

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5474f42
DSSWAT-5474_add_custom_transforms_on_the_fly
ivkrasovskiy Feb 9, 2023
3f3c96c
DSSWAT-5525_transformations_added
ivkrasovskiy Feb 18, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12,692 changes: 12,692 additions & 0 deletions experiments.ipynb
View file
Open in desktop

Large diffs are not rendered by default.

44 changes: 44 additions & 0 deletions heuristics/model/dataset.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,10 +2,49 @@
from typing import Dict, Optional

import pandas as pd
from avito_ds_swat_utils.storages.dwh import VerticaPandasConnection
from avito_ds_swat_utils.utils.image_manager import ImageManager
from PIL import Image
from torch.utils.data import Dataset
from torchvision import transforms

from .settings import (
IMAGES_DIR,
LABELS_PATH,
MAX_IMG_SIZE_STR,
)


def get_pd_dataset(cache_path=LABELS_PATH):
if cache_path is None or not os.path.exists(cache_path):
with VerticaPandasConnection() as dwh:
df = dwh.sql_to_pd('select * from dsswat.datasets_course_room_type')

df = df.dropna(subset=['image_id_ext'])
df['image_id_ext'] = df['image_id_ext'].astype(int)
df.reset_index(inplace=True)
if cache_path is not None:
df.to_csv(cache_path, index=False)
else:
df = pd.read_csv(cache_path)

return df


def load_images(image_ids, save_dir=IMAGES_DIR, img_size=MAX_IMG_SIZE_STR, image_save_shards=False):
image_manager = ImageManager()

image_manager.process_images(
image_id_list=image_ids,
schema='item',
size=img_size,
version=1,
private=True,
image_save_to_disk=True,
image_root_dir=save_dir,
image_save_shards=image_save_shards,
)


class TorchDataset(Dataset):
image_id_column: str = 'image_id_ext'
Expand All @@ -20,6 +59,7 @@ def __init__(
image_dir: str = None,
transformer: Optional[transforms.transforms.Compose] = None,
normalize_sample_weights: bool = True,
custom_transform : Optional[transforms.transforms.Compose] = None,
):
super(TorchDataset, self).__init__()
self.image_dir = image_dir
Expand All @@ -30,6 +70,7 @@ def __init__(
# self.df = self.df[self.df['result'] != -1]
self.transformer = transformer
self.image_cache = {}
self.custom_transform = custom_transform

def img_path_from_id(self, image_id):
path = os.path.join(self.image_dir, f'{image_id}.{self.img_extension}')
Expand Down Expand Up @@ -67,6 +108,9 @@ def __getitem__(self, idx):

self.image_cache[idx] = img

if self.custom_transform:
img = self.custom_transform(img)

# item = {
# 'img': img,
# 'label': item_series['result'],
Expand Down