Tim Esler comments

Results 14 comments of


                                            Tim Esler

Pre-Trained Models

@xiaoiker @chungyau97 there is a pytorch conversion of the two pretrained inception resnet v1 models (vggface and casia-web face) from davidsandberg/facenet here: https://github.com/timesler/facenet-pytorch Disclaimer: it is my repo.

MTCNN is failing to save cropped faces, but throws no errors.

I have not been able to reproduce this issue. Can you provide a complete, self-contained example?

Feature/add vggface2 labels

Fixes #60

box is a NoneType object

Hi @aakashg1999, does this only happen when there are no faces found in the frame?

mtcnn detect face error

@yueyihua can you let us know what versions are you using for: * facenet-pytorch * pytorch * torchvision * python

[Chatbot Arena] Add Falcon 40B model

@OAfzal could the random output be because fastchat is loading the model with dtype `float16` rather than `bfloat16`?

[Chatbot Arena] Add Falcon 40B model

@OAfzal I just saw that official support has been added for Falcon in https://github.com/huggingface/text-generation-inference, so you may be able to glean some insight there about how to get streaming working

memory leak in detect_face?

Does the memory accumulate with each loop iteration or is it a one off jump of 300MB?

How to train and return 128 embedding?

What do you need 128-dim embeddings for? It will probably be easier to either: 1. Add a single linear layer at the end of the model to do the mapping...

Add a `vertical_overflow='crop_above'` as an option to `Live()`

This feature would be amazing for CLI LLM chat apps!