Skip to content

Implementation of Face Embeddings using buffalo_l model#294

Merged
deven96 merged 11 commits intomainfrom
feat/buffalo-l-face-recognition
Feb 20, 2026
Merged

Implementation of Face Embeddings using buffalo_l model#294
deven96 merged 11 commits intomainfrom
feat/buffalo-l-face-recognition

Conversation

@deven96
Copy link
Copy Markdown
Owner

@deven96 deven96 commented Feb 16, 2026

No description provided.

@github-actions
Copy link
Copy Markdown

github-actions Bot commented Feb 16, 2026 •
edited
Loading

Test Results

264 tests   264 ✅  10m 47s ⏱️
 35 suites    0 💤
  4 files      0 ❌

Results for commit 5bc63a4.

♻️ This comment has been updated with latest results.

@deven96 deven96 force-pushed the feat/buffalo-l-face-recognition branch 2 times, most recently from f6499de to 976e797 Compare February 16, 2026 14:05
@github-actions
Copy link
Copy Markdown

github-actions Bot commented Feb 16, 2026 •
edited
Loading

Benchmark Results

group                                                        main                                   pr
-----                                                        ----                                   --
predicate_query_with_index/size_100                          1.00      3.0±0.00µs        ? ?/sec    1.05      3.1±0.00µs        ? ?/sec
predicate_query_with_index/size_1000                         1.00     33.2±0.03µs        ? ?/sec    1.01     33.6±0.03µs        ? ?/sec
predicate_query_with_index/size_10000                        1.02    403.9±0.83µs        ? ?/sec    1.00    394.5±1.42µs        ? ?/sec
predicate_query_with_index/size_100000                       1.07     10.5±0.91ms        ? ?/sec    1.00      9.8±0.85ms        ? ?/sec
predicate_query_without_index/size_100                       1.00      7.1±0.00µs        ? ?/sec    1.00      7.1±0.01µs        ? ?/sec
predicate_query_without_index/size_1000                      1.02    100.6±0.45µs        ? ?/sec    1.00     98.6±0.17µs        ? ?/sec
predicate_query_without_index/size_10000                     1.03    834.5±1.16µs        ? ?/sec    1.00    813.8±3.13µs        ? ?/sec
predicate_query_without_index/size_100000                    1.00     21.9±0.42ms        ? ?/sec    1.00     22.0±0.57ms        ? ?/sec
store_batch_insertion_without_predicates/size_100            1.00    234.5±2.07µs        ? ?/sec    1.00    235.0±1.22µs        ? ?/sec
store_batch_insertion_without_predicates/size_1000           1.00  1241.0±10.08µs        ? ?/sec    1.01  1253.5±11.01µs        ? ?/sec
store_batch_insertion_without_predicates/size_10000          1.00     14.2±0.28ms        ? ?/sec    1.02     14.4±0.18ms        ? ?/sec
store_batch_insertion_without_predicates/size_100000         1.00    135.6±0.87ms        ? ?/sec    1.00    136.0±0.70ms        ? ?/sec
store_retrieval_no_condition/size_100                        1.02    114.1±0.83µs        ? ?/sec    1.00    111.7±0.87µs        ? ?/sec
store_retrieval_no_condition/size_1000                       1.00   782.9±13.62µs        ? ?/sec    1.00    785.5±6.84µs        ? ?/sec
store_retrieval_no_condition/size_10000                      1.02      7.7±0.06ms        ? ?/sec    1.00      7.5±0.07ms        ? ?/sec
store_retrieval_no_condition/size_100000                     1.00     81.6±0.43ms        ? ?/sec    1.00     81.3±0.43ms        ? ?/sec
store_retrieval_non_linear_kdtree/size_100                   1.01    196.1±0.66µs        ? ?/sec    1.00    194.0±0.75µs        ? ?/sec
store_retrieval_non_linear_kdtree/size_1000                  1.00   1153.2±1.85µs        ? ?/sec    1.00   1149.3±1.75µs        ? ?/sec
store_retrieval_non_linear_kdtree/size_10000                 1.00     12.8±0.11ms        ? ?/sec    1.00     12.8±0.09ms        ? ?/sec
store_retrieval_non_linear_kdtree/size_100000                1.00    143.2±0.58ms        ? ?/sec    1.00    143.5±0.80ms        ? ?/sec
store_sequential_insertion_without_predicates/size_100       1.00    273.0±0.73µs        ? ?/sec    1.00    273.7±0.74µs        ? ?/sec
store_sequential_insertion_without_predicates/size_1000      1.00      2.7±0.01ms        ? ?/sec    1.01      2.7±0.01ms        ? ?/sec
store_sequential_insertion_without_predicates/size_10000     1.00     26.9±0.05ms        ? ?/sec    1.01     27.1±0.09ms        ? ?/sec
store_sequential_insertion_without_predicates/size_100000    1.00    269.5±1.42ms        ? ?/sec    1.01    272.1±1.18ms        ? ?/sec

derhnyel
derhnyel approved these changes Feb 16, 2026
View reviewed changes
Copy link
Copy Markdown
Collaborator

@derhnyel derhnyel left a comment •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Lgtm! CI is happy and no regressions though I didn't critically review this, I will let @Iamdavidonuh who has more context around the code give a critical review for this.

@deven96 deven96 force-pushed the feat/buffalo-l-face-recognition branch 2 times, most recently from da923f3 to c2473ef Compare February 16, 2026 17:46
@deven96 deven96 force-pushed the feat/buffalo-l-face-recognition branch from c2473ef to 01e1453 Compare February 16, 2026 19:49
HAKSOAT
HAKSOAT reviewed Feb 18, 2026
View reviewed changes
Copy link
Copy Markdown
Collaborator

@HAKSOAT HAKSOAT left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@deven96 I added some comments.

I'd like to read the retinaface paper. So will review that part more after I do so.

Comment thread ahnlich/ai/src/engine/ai/providers/processors/imagearray_to_ndarray.rs Outdated
Comment thread ahnlich/ai/src/engine/ai/providers/ort/models/buffalo_l.rs Outdated
Comment thread ahnlich/ai/src/tests/buffalo_l_test.rs
@HAKSOAT
Copy link
Copy Markdown
Collaborator

HAKSOAT commented Feb 18, 2026

The insightface model card says:

🔔 ALL models are available for non-commercial research purposes only.

https://github.com/deepinsight/insightface/tree/master/model_zoo

This is something to consider for this integration. Either have a warning or disclaimer attached to the model.

@deven96
Copy link
Copy Markdown
Owner Author

deven96 commented Feb 18, 2026

The insightface model card says:

🔔 ALL models are available for non-commercial research purposes only.

https://github.com/deepinsight/insightface/tree/master/model_zoo

This is something to consider for this integration. Either have a warning or disclaimer attached to the model.

Oof that's tough... we can add the warning and have folks obtain the license themselves

@HAKSOAT
Copy link
Copy Markdown
Collaborator

HAKSOAT commented Feb 18, 2026

Oof that's tough... we can add the warning and have folks obtain the license themselves

Yeah. I think it's quite discouraging. I'm not sure how obtaining the license might work, but a warning would be good.

See:

@deven96
Copy link
Copy Markdown
Owner Author

deven96 commented Feb 18, 2026

Oof that's tough... we can add the warning and have folks obtain the license themselves

Yeah. I think it's quite discouraging. I'm not sure how obtaining the license might work, but a warning would be good.

See:

Alright then I will add the warning and make a note even in public facing docs

@deven96 deven96 force-pushed the feat/buffalo-l-face-recognition branch from 085462f to 725a074 Compare February 18, 2026 14:26
@deven96 deven96 requested a review from HAKSOAT February 18, 2026 21:08
@deven96
feat(ai): add SFace+YuNet face recognition pipeline (SFACE_YUNET = 8)
544aec3 
Commercially-licensed alternative to Buffalo_L using YuNet (MIT) for face detection
and SFace (Apache 2.0) for 128-dim embeddings. Extracts shared face_align module
(NMS, warp_affine, similarity transform) used by both pipelines with minor perf
optimisations. Regenerates Go, Node, and Python SDK clients.
HAKSOAT
HAKSOAT approved these changes Feb 20, 2026
View reviewed changes
Copy link
Copy Markdown
Collaborator

@HAKSOAT HAKSOAT left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Checked the changes. Looks good too.

@deven96 deven96 force-pushed the feat/buffalo-l-face-recognition branch 8 times, most recently from aa1f12b to 56c3ab4 Compare February 20, 2026 14:20
@deven96 deven96 force-pushed the feat/buffalo-l-face-recognition branch from 56c3ab4 to 5bc63a4 Compare February 20, 2026 14:32
@deven96 deven96 merged commit 7fdb808 into main Feb 20, 2026
6 of 7 checks passed
@deven96 deven96 deleted the feat/buffalo-l-face-recognition branch February 20, 2026 14:45
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@HAKSOAT HAKSOAT HAKSOAT approved these changes

@Iamdavidonuh Iamdavidonuh Iamdavidonuh approved these changes

@derhnyel derhnyel derhnyel approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@deven96 @HAKSOAT @Iamdavidonuh @derhnyel