# FaceNetTRT Face Encodings

**URL:** <https://forums.developer.nvidia.com/t/facenettrt-face-encodings/107882>\
**Category:** Jetson Nano\
**Created:** [November 22, 2019, 3:56pm UTC](https://forums.developer.nvidia.com/t/facenettrt-face-encodings/107882 "2019-11-22T15:56:49Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![yomuevans](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/yomuevans/32/19784_2.png) [@yomuevans](https://forums.developer.nvidia.com/u/yomuevans)\
**Post date:** [November 22, 2019, 3:56pm UTC](https://forums.developer.nvidia.com/t/facenettrt-face-encodings/107882/1 "2019-11-22T15:56:49Z")

</div>

I just realised Jetson Inference repo has a facenet mode, of which performs really well on the jetson nano, but all it does is face detection [GitHub - dusty-nv/jetson-inference: Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson.](https://github.com/dusty-nv/jetson-inference) . I was hoping to get face embeddings as implemented by [GitHub - davidsandberg/facenet: Face recognition using Tensorflow](https://github.com/davidsandberg/facenet). Thanks. would really be helpful

---

<div class="post-metadata">

**Author:** ![NVES\_R](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/nves_r/32/14043_2.png) [@NVES\_R](https://forums.developer.nvidia.com/u/NVES_R)\
**Post date:** [November 22, 2019, 7:14pm UTC](https://forums.developer.nvidia.com/t/facenettrt-face-encodings/107882/2 "2019-11-22T19:14:25Z")

</div>

Moved to Jetson Nano forum.

---

<div class="post-metadata">

**Author:** ![dusty\_nv](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/dusty_nv/32/14043_2.png) [@dusty\_nv](https://forums.developer.nvidia.com/u/dusty_nv)\
**Post date:** [November 22, 2019, 7:38pm UTC](https://forums.developer.nvidia.com/t/facenettrt-face-encodings/107882/3 "2019-11-22T19:38:12Z")

</div>

Hi yomuevans, it’s a caffe model, so you could look through the facenet prototxt for the layer you are looking for, and specify that as an output layer in the code. You would need to manually modify the detectNet code to expect the output of this additional layer.