Skip to content

Commit 2cba4a1

Browse files
committed
Moved Face-data-migration.md updates to docs
1 parent beba65d commit 2cba4a1

File tree

2 files changed

+29
-66
lines changed

2 files changed

+29
-66
lines changed

Face-data-migration.md

Lines changed: 0 additions & 44 deletions
This file was deleted.

docs/Face-data-migration.md

Lines changed: 29 additions & 22 deletions
Original file line numberDiff line numberDiff line change
@@ -2,36 +2,43 @@
22

33
## When do you need to migrate data
44

5-
When you upload a new known image to the Face Collection,
6-
CompreFace uses a neural network model to calculate an embedding (also known as face features),
7-
which is basically an array of 512 or 128 numbers.
8-
Then CompreFace saves it to the database to use in the future comparison - when you upload a face to recognize,
9-
CompreFace again calculates an embedding and compares it to saved embeddings.
10-
11-
The important thing here is that every neural network model will calculate different embeddings.
12-
It means that it makes sense to compare embeddings calculated by the same neural network model.
13-
14-
CompreFace doesn't change the neural network model during its work, so normally you don’t need to migrate your face data.
15-
If you want to try [custom build](Custom-builds.md), be very careful - look at the table [here](../custom-builds/README.md),
16-
column `Face recognition model` - if the model changed, you need to run a migration.
5+
When you upload a new known image to the Face Collection, CompreFace
6+
uses a neural network model to calculate an embedding (also known as
7+
face features), which is an array of 512 or 128 numbers. Then CompreFace
8+
saves it to the database to use in the future comparison - when you
9+
upload a face to recognize, CompreFace again calculates an embedding and
10+
compares it to saved embeddings.
11+
12+
The important thing here is that every neural network model calculates
13+
different embeddings. Therefore, it means that it makes sense to compare
14+
embeddings calculated by the same neural network model.
15+
16+
CompreFace doesn't change the neural network model during its work, so
17+
you usually don't need to migrate your face data. If you want to try
18+
[custom build](Custom-builds.md), be very careful - look at the table
19+
[here](../custom-builds/README.md), column `Face recognition model` - if
20+
the model changed, you need to run a migration.
1721

1822
## Limitations
1923

20-
If you run CompreFace in the [“not saving images to database” mode](Configuration.md)(`save_images_to_db=false`),
21-
you won’t be able to migrate data as the original images are required for migration.
24+
If you run CompreFace in the ["not saving images to database"
25+
mode](Configuration.md)(`save_images_to_db=false`), you won't be
26+
able to migrate data as the original images are required for migration.
2227

23-
The only solution here is to delete all images from Face Collection and upload them again.
28+
The only solution here is to delete all images from Face Collection and
29+
upload them again.
2430

2531
## How to perform a migration
2632

27-
Current migration was written for internal usage and wasn’t tested enough, so please do a backup copy of the database and perform migration at your own risk.
33+
Current migration was written for internal usage and wasn't tested
34+
enough, so please do a backup copy of the database and perform migration
35+
at your own risk.
2836

2937
REST request to start migration:
3038

31-
```shell
32-
curl -i -X POST \
33-
'http://localhost:8000/api/v1/migrate'
34-
```
39+
curl -i -X POST \
40+
'http://localhost:8000/api/v1/migrate'
3541

36-
This rest endpoint is asynchronous, it will start the migration and return a response immediately.
37-
To understand if the migration is successful, please look at logs for “Migration successfully finished” text.
42+
This rest endpoint is asynchronous; it starts the migration and returns
43+
a response immediately. Please look at logs for "Migration successfully
44+
finished" text to understand if the migration is successful.

0 commit comments

Comments
 (0)