Memap dataset for multimodal data #47
Conversation
…tests for image and multimodal data
|
@ndoll1998, please make sure to enable the pre-commit hooks. |
src/modalities/dataloader/codecs.py
Outdated
| @@ -0,0 +1,111 @@ | |||
| from abc import ABC, abstractmethod | |||
| from typing import TypeVar, Generic, Optional, Dict, Any | |||
src/modalities/dataloader/codecs.py
Outdated
| # write image to buffer | ||
| with Image.open(img_file_path) as img: | ||
| img.save(buf, format=self._format) | ||
| # retuen buffer content |
| from typing import TypeVar, Generic, Optional, Dict, Any | ||
|
|
||
| from io import BytesIO | ||
| from PIL import Image |
| index = [(4, 24), (28, 40), (68, 12), (80, 4)] # [(index,len), ...] -> in 4 bytes #lengths: 6,10,3,1 | ||
| data += codecs_bytes | ||
| # index | ||
| index = [(16, 24), (40, 28), (68, 12), (80, 16)] # [(index,len), ...] -> in 4 bytes #lengths: 6,10,3,1 |
There was a problem hiding this comment.
I am unsure if this index is correct. I think it should be index= [(16,24), (40,40), (80,12), (92, 4)]. Maybe I have not understood the concept. Could you please clarify @ndoll1998
There was a problem hiding this comment.
The specific values of the index shouldn't matter as long as they follow the schema (begin, length). Both the index lists do so.
Take for example the following index item (16, 24). This index refers to the sequence [0, 1, 2, 3, 4, 5] as it begins at the 16th byte right after the header bytes and has a length of 24 bytes <=> 6 integer values.
Given that this is just a dummy example, the specific sequences in the dataset are not of interest.
le1nux
force-pushed
the
main
branch
3 times, most recently
from
cb6e816 to
179052b
Compare
|
Will this PR be continued or can it be closed? @spravil @ndoll1998 |
No description provided.